{"id": "v3_3efd65b51fd6b9f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.31, 0.63, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_719882f82c0d1749_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ffd73d6f3643cd7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.16, 0.69, 0.12], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6002eeb5233dee24_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba1e862ebb5fa056_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.24, 0.53, 0.15, 0.01], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de606aa6ae67d732_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.53, 0.29, 0.05, 0.13], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5301f774082709b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d117fb9c4159689_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.41, 0.49, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fc76ef3c9e8924a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.17, 0.53], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a8ca0ee276120ae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.05, 0.52, 0.41, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44c2397332e5ce85_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8264cc8dcae09ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54c97a589689f9f1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.29, 0.32, 0.26], "state": "An automation plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_facd826ee4d3b772_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d271b18350ba6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.23, 0.24, 0.36, 0.07, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3979472bd68dbc6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.25, 0.59, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_234a075eae0575ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.23, 0.45, 0.25, 0.02], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43b93adcf8784411_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c17837781fd0663_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0a0489ce3686a4f_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_977f892687670541_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccfaf7f9212bdf92_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.75, 0.08, 0.14], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab4e4a897a07f62e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.24, 0.1, 0.52], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a57547febd908917_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72c1790e619df6be_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.08, 0.91, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c4444912d2b283a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.43, 0.54, 0.01], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f53bf6e414d7c816_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.18, 0.58, 0.19, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e8cb07fbef09878_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.6, 0.27, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0962459101d9e7c1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54d5fa1b307e9820_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.26, 0.18, 0.33, 0.08, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac9e89a5474bcd6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.56, 0.1, 0.31], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_484c6c5fd7f02d10_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.03, 0.28, 0.6, 0.04], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f15d697611568c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2223a944aeb00b0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.1, 0.79, 0.1], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2dc2e07037e910c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.27, 0.6, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fbd5ea0628b7444_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.77, 0.14, 0.08], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72580fd4b4543676_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.31, 0.43, 0.19, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ba37c74bed37a0d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.18, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c43b254773c18918_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.15, 0.7, 0.09], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14a01012c8c330e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9333317779ec4a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9b1e527b5b4d995_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.25, 0.17, 0.36, 0.13, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da7ec481091c60ab_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b558e7427c84d58a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.1, 0.7, 0.15], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_919ed2d2f3055219_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.5, 0.43, 0.01], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ce4856ecb6dd680_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62256a278a065735_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bed8d226119c25b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.76, 0.19, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6de91eb756d4eaee_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.53, 0.37, 0.09], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5286466a7dbbbdf_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34a6a004792cf819_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.36, 0.55, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0de791aa3609e92e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.31, 0.56, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e061b301f3f14733_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.13, 0.41, 0.4, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b421120af978f63_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.05, 0.41, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45d65f9f52d26978_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.97, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eb33ee004eaddc34_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.21, 0.12], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dff841a5fca40b79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.04, 0.12, 0.3, 0.4, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54130c1ed1ed902b_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_713fdde29303229a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.65, 0.06], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcbfc548f1f66d1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.25, 0.6, 0.05], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9dda360e0fd8f05_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.24, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6e43fe666e20628_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.43, 0.09, 0.38], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f49866d5c9129a42_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1561a885ffa801fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.02, 0.11, 0.61, 0.18], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7484b95f2950c9f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.04, 0.28, 0.58, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f77a9a22fb2c3429_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0202, 0.0505, 0.2121, 0.5657, 0.1111], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2199ea34c82ddbfc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.5, 0.13, 0.27, 0.1], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39919b1ff3e12e7e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.31, 0.18, 0.31, 0.06, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c9e90d4ff10c448_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.29, 0.12, 0.33, 0.12, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf55c73888d068ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.53, 0.32, 0.02], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab40858c20e66da9_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_834929b62c6a1bdf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.17, 0.22, 0.16, 0.34, 0.09, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4f82fbac181400a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.19, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c757b0e72c2fa035_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.08, 0.46, 0.34, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9dd70c40a315caa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.03, 0.08, 0.32, 0.43, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c2fbe5cd2f83545_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.17, 0.17, 0.36, 0.13, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ba7ad1b503e9233_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbfb49fc3e225d22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.77, 0.14, 0.09], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5ca8198862eba52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.2, 0.6, 0.12], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e1d13d0afdf691_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cdb58f1055fa3c8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.23, 0.61, 0.11, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6af9e9709bc33d_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cb8dccb3461dc0f_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_618dbd980e02fab5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.95, 0.04], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8cefde14f961d9e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.04, 0.04, 0.89], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fa17a4512fca752_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.31, 0.13, 0.33, 0.1, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f29b86096a51f068_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.03, 0.93], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_700ecd50bb38daa8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.34, 0.53, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ac6e3cf729a318c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.32, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c20b6926910828f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.13, 0.13], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a98431bcea23e4dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.06, 0.54, 0.33], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abb7db2cc34ec3b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6bca7419a25f7c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.14, 0.49, 0.3, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d03f29a753f8f9ea_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ffd8aef1ff27eb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.12, 0.47, 0.27, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83ba125d5543f087_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca8dd4cc0c1619dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.32, 0.6, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16fd45c9d5e925a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02ef1aee0da9f8fa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.44, 0.29, 0.21], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8074deb77843f9db_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb4aeceec6a7dc1d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.16, 0.82], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b7eb9f943b36bc6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.15, 0.37, 0.33, 0.04], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca2b45eb62a0d27_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.89, 0.1, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32b8f50877956722_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.71, 0.1], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_738eb66f1f2166ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_609ff41c5b4fe96e_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52fd070b98f08e6a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.74, 0.1], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff0946e7498a5dcc_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_610e3861aa143625_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03c241c77628042a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.73, 0.16, 0.08], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be4c0d9b40bef977_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.33, 0.47, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a5b45d77bb52365_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.25, 0.17, 0.31, 0.07, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d0f5c016f621589_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.77, 0.07], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43cb5f8a9aa25958_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.08, 0.46, 0.39, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6579746b10a4644_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0101, 0.0505, 0.3232, 0.5556, 0.0202], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a059ba9c436ff76_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.68, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa82e22dd8820be6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.15, 0.71, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98721a84bc59f172_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.09, 0.18, 0.44, 0.16, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acd9b1b222c524b4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.94, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_79dfcb3a99a60101_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.86, 0.08, 0.04, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98d1c621689639c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.2, 0.19, 0.38, 0.11, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942106c6de437eab_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd441ac060bc3d0d_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afb0412a7d68524c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.49, 0.37, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6e8e876991ecf8a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.15, 0.43, 0.33, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e87150e41aac1185_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.03, 0.55, 0.37], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_742a70b645503ecf_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b3b8bc63e89be5c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.1, 0.37, 0.41, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_639f06566c041fa2_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d466ad6788286eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9140f6ac51860c06_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88bcb9ec5fad216a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.61, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e54d9e1bfb2e998_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.39, 0.47, 0.02], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_362a4b974ba77252_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.48, 0.33, 0.02], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b8b19879c12932c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.52, 0.04, 0.42], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26dbbeb886f3f1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.41, 0.47, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457b7ea1440e0296_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.16, 0.17, 0.42, 0.14, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d857804cdfc5c101_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.21, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76504429f57f0da7_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6b8f04c6d95bc6e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.06, 0.24, 0.21], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_451ff08e8fd9b65b_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad062133d0c325ea_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d1b3d91e66e236d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.31, 0.5, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e228a5203bbec4_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48d8218c00f450a7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.31, 0.2], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f768660d93aaf24_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.13, 0.46, 0.28, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63d5277473509627_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2a2d83a49628ee2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.44, 0.53, 0.0], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bf312ca3616442b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.84, 0.13], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efc16bb63a7b65f5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.72, 0.12, 0.13], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b58857e1e41d99c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae350bfe85dd0516_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.36, 0.47, 0.14, 0.01], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eab9bd675bbbdc79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.14, 0.31, 0.49, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1140993bdd06b32_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b99007e3a536ffa9_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09ede11f5a8fa5bb_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f77a9a22fb2c3429_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92f4fe645926f762_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.34, 0.66, 0.0], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94571f9e2128dee7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.08, 0.14, 0.48, 0.24, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_795285a6a5af3871_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6fc617bfc967933_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff30d71d81f121bb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.49, 0.23], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57b1b3ff7615dc4d_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5de1114bd9c12cc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.15, 0.37], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ddd2ebb8f7f63a2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.29, 0.59, 0.04], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8ce88e75a3459b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5498145fe6a54907_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.4, 0.2, 0.27, 0.04, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7bb9036c6ffd8030_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.74, 0.2, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90b951b396fe70c9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.12, 0.39, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0b21bfc4172c472_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.16, 0.14, 0.4, 0.22, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87802b8d9fdc10d1_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24edde6920904b2c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_072d65f96448fcce_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_236be50aa8a0bd05_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acbfbb8c105d2748_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.67, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b46507de75ae98eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88d9e157ca858567_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a9d2209b76b8fe1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.16, 0.21, 0.42, 0.11, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dc2429fd1686d79_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cf3ffe988217525_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.19, 0.4, 0.2, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0ae71b367d3feca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.12, 0.24, 0.38, 0.13, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f05335e5cb4a6fce_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.33, 0.08], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9fdafc77124012f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.51, 0.43], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b56d080ec1e3f63_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.14, 0.52, 0.21, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c700f379d568ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.12, 0.4, 0.4, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ab8a43409dd8c89_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.32, 0.07], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7a9557c57aaf356_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.37, 0.22, 0.15], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_573828a2a0359a96_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.05, 0.16, 0.54, 0.18], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b8b19879c12932c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9ebfaed888061d0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.36, 0.54, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6231276c852ae808_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.5, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f566ae45ea7d0c_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8f7bb9161a36d30_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d87fdbdc1a2c90f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.21, 0.48, 0.2, 0.01], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d6e350c427ae30a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.04, 0.23, 0.63, 0.1], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e04db42af010212_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d4ba213ce853c18_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33d3dd2a6643c3f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.26, 0.56, 0.08], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_811d577aa5227c54_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.39, 0.48, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_519819b6824bfe01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.27, 0.54, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9488f56f4cbc6921_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.4, 0.16], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea021e551d8aa539_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12f390717a0585ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.21, 0.79], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_442f16a6d9a0f15e_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ee15a08a5c44163_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.16, 0.68, 0.09], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d291ebf8902aec7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.11, 0.11, 0.42, 0.25, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13ad471089a036bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76089f4abe448de6_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0104129156e13a82_n", "kind": "noul", "options": ["false", "true"], "target": [0.13, 0.87], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e6d1bfeed21de41_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97cc7ba3218e88ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1f7ca004217b52b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.46, 0.45, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05e81f95f90fa438_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9039aef5521f6beb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.77, 0.12, 0.11], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c3ccaf2ab5cdc68_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c3f04724692fb0b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.57, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ddc278c3c0f1ddfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39d9f6a46c868969_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d06fd2a0509d26f0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.15, 0.61, 0.24], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc8f2bb182152175_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.66, 0.21], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d78e27fc451a3646_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.61, 0.1, 0.17], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd357df2882a04a6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.18, 0.03, 0.64], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e5b2eaa803a5c77_n", "kind": "noul", "options": ["false", "true"], "target": [0.24, 0.76], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe8e23a9fc1bbeca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.21, 0.16, 0.35, 0.1, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c3fd4760f62c226_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b45a658e0ec00ad1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.11, 0.77, 0.09], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6933df6e028c48e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.23, 0.66, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b124407b91a3a718_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.32, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1715ee6cf1bebd6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0101, 0.2828, 0.6667, 0.0404], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff0946e7498a5dcc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.81, 0.13, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_437894c57f111804_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.12, 0.5, 0.34, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bd7ce93aa5572d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.09, 0.27, 0.41, 0.2, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51cb60461e068ac2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.71, 0.2, 0.08], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d7d984399acf0e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.12, 0.83], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b4539bb10a44f15_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.37, 0.08, 0.32, 0.1, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de0e925a5a4ea8fe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.78, 0.2, 0.02], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c309cbd215074a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce3f0d0ab4125653_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9120c1cc59c1cb9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d00f6a757531e258_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae81b690c8ac3667_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.06, 0.76, 0.15], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fad105038a580ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.46, 0.21, 0.23, 0.01, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bdd1a9c908dfdf6_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe6f42ea1e457817_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.65, 0.01, 0.26, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e7bbd69c2e4aeea_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.06, 0.53, 0.27], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d2fb9d037d37285_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e8131ec88b99f99_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.28, 0.14, 0.36, 0.05, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1ae8c481e775f7b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.38, 0.02], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_361e46db782ef7b6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.27, 0.06], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b4dac0b8f1fce94_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34ae804128c31609_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.27, 0.45, 0.17, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f12e76a8a4df476_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.24, 0.65, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eee63d3463464fe7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.02, 0.06, 0.16], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d75ff1e736c4a8a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.09, 0.17, 0.46, 0.22, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49109cc752fa4365_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_103691a9d398a011_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.08, 0.75, 0.17], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e570fa9aa32ad948_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac1b5f3ac7841ec1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.02, 0.05, 0.28, 0.51, 0.05], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b9bc07380cc09b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d46b62e22f01c00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.29, 0.14, 0.36, 0.1, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55abc08beec17438_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91255cdbd80c2f7d_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_729cb0865cfc328c_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c44c4c9474fbf9a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.22, 0.19, 0.42, 0.06, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b946cfc07f7090f_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_888d3188e4b47d80_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d530d2fec94d0a33_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a59d3fef0d1a6b33_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.48, 0.09], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c251f31c548527da_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e334abc21f31583_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.81, 0.15], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b9290268ccf3f58_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a829c511b6f1d38b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.1, 0.31, 0.47, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56108d3783f302f2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.93, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3184e6a0424a5d9f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.92, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b22b09b9e1ca5ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e67e0c5a2fc3bf6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.04, 0.22, 0.6, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a39c2c134fbbdea_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.36, 0.01], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_939206d6ffe7eb9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e968dc3cb4f4db4_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f02eea3042987e21_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6107a1159d65aba_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dd8ed7513e8c443_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3c2ea197d6dd608_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_003d321459de0e7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.03, 0.11, 0.5, 0.27], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf5d267550558a3_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0334399392b60f51_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_642cb90838e29889_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f87e544e6935872_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.42, 0.38, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31f9f37481d2d97a_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc5e5ad4dca3ee2c_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9c96977ff33f5a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d291ebf8902aec7_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccedb3e74554e314_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.13, 0.75, 0.1], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abf6a0508dde06e9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.27, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43d1a96eb9414e7a_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1d23118b7ba63ad_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.29, 0.21], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4ec5311c91a1417_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35c4096348653c21_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.49, 0.4, 0.09], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c4fecb78131315d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.19, 0.52, 0.19, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dde758f62343edd_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a31c898509ec7baa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.32, 0.16, 0.35, 0.08, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a98790343f971b1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.1, 0.19, 0.48, 0.2, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13be773c0297e86d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.53, 0.13], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_961a2f4afce49993_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.13, 0.42, 0.34, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c69086e804f4de1f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.37, 0.13, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e67f41eef1c700b0_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5333fa36f9f7007_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab4e54aaacf5bd2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a46c35fba63812d0_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66367726b2399f72_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21ef33cbaf481fc8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.07, 0.16, 0.43, 0.22, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b472344d7024138_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0505, 0.1414, 0.5455, 0.2323, 0.0101], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d530d2fec94d0a33_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.19, 0.55, 0.16], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e33940d3dbb79e0b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.1, 0.35, 0.47, 0.04], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a8e6f8c64e2ff8c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.52, 0.28, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aed0b63e5c03d329_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90db561c14637f6a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74d59ce47911ead2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.17, 0.1, 0.66], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69fd5a212438894f_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7afdf61d85fad916_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b61d59616769564b_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b22b09b9e1ca5ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.5, 0.32, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b2039adbd28ea39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.1, 0.52, 0.35, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aa4cdc8551e28d7_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a6cea40d3ef1f48_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b025377bb7ed6ede_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.39, 0.57, 0.02], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df4853dcb77320ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17ad7138867e11b9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.5, 0.4, 0.08], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f4253883d3e6d55_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5756a2510700b749_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.73, 0.26, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0113d0323359c538_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.28, 0.62, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ac85f8a4234dedf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.85, 0.03, 0.05, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42e7d70f881666e1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.61, 0.1, 0.26], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a4427b786583342_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.46, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29fa85b639d7e3f5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.23, 0.61, 0.06], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_facf915526833856_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.84, 0.12, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3978ec1143b544e_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4966af296c2b0d08_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.54, 0.44, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7894accfe1731fab_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.38, 0.0, 0.01, 0.61], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da1a7e66012e69a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.18, 0.51, 0.2, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b89f315163b792d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.21, 0.57, 0.15], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc116055dbad736f_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d46b62e22f01c00_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d10870c3ba3ec1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.1, 0.38, 0.39, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21074d093bae630_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.03, 0.96], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de3c3ab8a593f8d5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.22, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1353a18bc0141bd3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.56, 0.08, 0.34], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c04838e8b5fb06ef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.13, 0.21, 0.44, 0.08, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fe822e47a9b6628_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_014932068facba82_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.9, 0.01, 0.03, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2d8b03086a5a292_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c79b84590c24c1d_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_644841e296a23901_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.41, 0.53, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_748d115ea9d44aef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.43, 0.44, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6923c78ce5427ab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.22, 0.58, 0.1], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8686d3ec4bc7cdb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.01, 0.05, 0.89], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef0ac2a7489dfa66_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.08, 0.74, 0.18], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cfcf85c28948c68_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.79, 0.15, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a09946e4e742354_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.24, 0.51, 0.2, 0.01], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be7c71801ea27565_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3abd9133fc74e47a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.18, 0.18, 0.42, 0.09, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f36e3f3387d1bcc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.54, 0.03], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_463fd86d4d16bd42_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.43, 0.27, 0.1, 0.2], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fdac7ababff7397a_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8cc9324fc7e00a5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.35, 0.53, 0.1], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b1ca49f3e6c5264_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.68, 0.28, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87bc98fedd0d714a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.55, 0.42, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8e2c13db052f0a4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.22, 0.06, 0.35, 0.37], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85511f46a92de3ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.41, 0.06], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b15d0993854f063_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.31, 0.29, 0.39], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d383dcd0814b7642_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.31, 0.12, 0.28, 0.07, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6561aa7f547dce25_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.32, 0.06, 0.3, 0.12, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5684d44875c68eb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.11, 0.58, 0.29], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_900b95f30c0ca7e8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0202, 0.0606, 0.2727, 0.5455, 0.0707], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75fbf17ed50a98e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.03, 0.24, 0.6, 0.03], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e54d9e1bfb2e998_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.75, 0.22, 0.03], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c5218b87ea0fb89_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.28, 0.54, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd11ff0c5bb4dc35_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_825f1de029b6647e_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12f390717a0585ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.37, 0.51, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97ffbc1ddf4856ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.03, 0.95], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48b0e98a4a66b49c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.13, 0.7, 0.15], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eb14c7aae97de9c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.29, 0.13], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29e1dbe1671bdd21_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_566c7b9721627609_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.83, 0.11, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15420d877f663596_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6087a465590c140b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.79, 0.1], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91506538ffecf3de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.48, 0.37, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86290f0a4f672e36_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.51, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e29f1286a552c7b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.86, 0.09, 0.04], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ddaba8510144067_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.18, 0.48, 0.22, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e99fb5d682e74a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.23, 0.22, 0.48], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a01b96f27bfc0b29_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e43c31528acbc00_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4be6360aa7923d8b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.31, 0.51, 0.04], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5ea568950478a00_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.62, 0.36, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5d004d79878e53_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.36, 0.59, 0.01], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e4cf06d1de55e3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.02, 0.96, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a58ec7a7bfcbd56c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.24, 0.5, 0.13, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1896a282d360f25a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.12, 0.54, 0.29, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a39c2c134fbbdea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.55, 0.42, 0.01], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d7992c70c9b9475_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.88, 0.08, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08aa7edfc96aa080_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d93c13872898063_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.64, 0.13], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21543e68a54acece_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.22, 0.67, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da9d31fdc7bb7e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10c19d5ca7961864_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b04bdcb096a65ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2a62b47d5879b03_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.58, 0.22, 0.09], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1772c9ab9fae2a96_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.86, 0.01, 0.1, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_443d27d62a8efe06_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8809a18ce1d14791_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1306acffc48aefb0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53aa6f13658fd53c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.32, 0.29, 0.34], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e07f4b429ed48587_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7384579268b43d1_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fd1e0ac8834f163_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0101, 0.1616, 0.6869, 0.1313], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01da22fd7035ddcc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.66, 0.3], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a75ef54655d246cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.48, 0.43, 0.01], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec1408e7cb758df_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.18, 0.33, 0.08], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d3d3cb81ff91d34_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.17, 0.72, 0.1], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f6482186d61d23a_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eab9bd675bbbdc79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.6869, 0.0202, 0.2323, 0.0606], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea65180d8b679f93_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f028c40ab18d8ad5_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96b1c32131d4b147_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7504ae3bf5d8443_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e86a0f4d6dd519e4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.48, 0.01], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_435f2a00676a423c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.03, 0.01, 0.58], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9c96977ff33f5a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.24, 0.51, 0.22, 0.01], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a39f9b995d196004_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.3, 0.68, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e20f8d372d4969b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0101, 0.0101, 0.1111, 0.6869, 0.1313], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a98431bcea23e4dd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.09, 0.84, 0.07], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12b22ed7fc5446b5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.03, 0.2, 0.63, 0.06], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26a1f0dfd25aa931_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.12, 0.37, 0.39, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_718a22a42ac21bcb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.85, 0.01, 0.05, 0.09], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43c4c231c0b289fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62256a278a065735_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.7, 0.24, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a7cde69c202d5fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1b18fefa2c73a1d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.49, 0.18, 0.2], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa072517cc2732c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2094f915535298e0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.36, 0.18, 0.28], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c1bd6823668fbb2_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e6aa0f03e64550f_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3169b463c94a8358_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.28, 0.02, 0.69, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fff8d5e8907ea1cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f6cb6c042d1a44c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.35, 0.56, 0.03], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3367a439cb75e35a_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3d67b8a93aff9f1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.43, 0.49, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4d1859cfa174a01_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.11, 0.56, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b421120af978f63_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_554dc6f0531144bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.19, 0.51, 0.17, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b536be7294042d8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.24, 0.02], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff7ec5ffebc09d57_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41818955047ed9ce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.57, 0.28, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d2fb9d037d37285_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.67, 0.08, 0.21], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e160dca304097d_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfecae896770d57f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3, 0.08, 0.55, 0.07], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2864e21eea12f295_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.98, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6f3e0eaaae957d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.07, 0.17, 0.38, 0.23, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42eec61568828ed3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.52, 0.46, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_622cd3abd9843a56_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.56, 0.25, 0.13], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b13101125ef6e21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.05, 0.78, 0.17], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c44f2f63cddcc8d_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db76fee0b259e6cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.08, 0.42, 0.41, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdc4890f5e7f634a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.67, 0.15, 0.13, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24090441f34b8ae1_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6cad2b588c6f69_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.11, 0.49, 0.35, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88e3fb83c65d665d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.25, 0.37, 0.04, 0.24, 0.07, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53e11686442fc94f_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc96f4f36fd692b3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.35, 0.6, 0.02], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34a6a004792cf819_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.37, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_983264d1eb601741_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.37, 0.53, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d91fb78fc504c763_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.19, 0.6, 0.11], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bc661d3a4234258_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52524896ae76d0dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e264baf07527559_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14e94b5daf0177db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.2, 0.39, 0.21, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ab0765156c68db9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.23, 0.04], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25a498a7bc050295_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e64cf2928acde53c_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ace53cfc46edd7a_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a1d28c0999ed486_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.11, 0.37, 0.49, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a403932a8c09a3e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.45, 0.48, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7449dedabcd20767_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.12, 0.4, 0.32], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a1d8e68b8596943_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54c228115ecee8b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.41, 0.45, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc533a7576106b3c_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6216ca71d1f2dd5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.51, 0.07, 0.3, 0.12], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_478b6646136138e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.43, 0.47, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d892f2b26a15a333_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3984b2c495c99fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.29, 0.6, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37795e7c34811c1c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66fd83b38be1e340_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5333fa36f9f7007_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.81, 0.17, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98c01fa5c0ae5ba5_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18547896f5199fd4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.3, 0.56, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c54e23bc2145736_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.07, 0.71, 0.18], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d5c8a702c24596_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.69, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d4ba213ce853c18_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.25, 0.37, 0.07, 0.31], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_361e46db782ef7b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.28, 0.62, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1cd9e71d4018ede_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_924e149ba2fd20d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.21, 0.18, 0.37, 0.09, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86ed3d98659418b7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.0, 0.98], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8d76913dc039378_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.26, 0.64, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a9b344c3038a211_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.76, 0.13], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_983264d1eb601741_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.53, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47ea5da8dfab6b90_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.17, 0.14, 0.66], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_176b4cc1beae9130_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b1ca49f3e6c5264_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41d9d6d888d9a6e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a696407522a9bc64_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.11, 0.86, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a2b56812de9a4a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66ac27615660abda_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.4, 0.34, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12b6d3a797355ef2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.22, 0.64, 0.11, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3ebd40a14cfa4e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7cfb391d70e902f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.3, 0.46, 0.07, 0.0], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6377eb523424c5e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37795e7c34811c1c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.14, 0.13, 0.72], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fcc88289efa28377_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.45, 0.05, 0.22, 0.07, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8dec5643dba879bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9a8789150f5d774_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.47, 0.5, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14594dbf022e9748_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71f93bc0bd3a0c13_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b4dacde4eba41b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1531e3b7afb2e3c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0202, 0.2222, 0.5657, 0.1818, 0.0], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53b6608d177dc514_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7899499ff84a8e85_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.6, 0.14, 0.23], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5561e740ebf0cd4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.13, 0.4, 0.37, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_631efd049871575d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.47, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eef4538f87ee3734_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.1, 0.55, 0.28, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb02f5581ae8fb74_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f4ac2c7e18f944f_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c36160a175048ec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.75, 0.05, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6592839acac290a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.35, 0.0], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df609e0542c72381_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.16, 0.46, 0.35, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6555216168a7a868_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7be40e18def55d85_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23208506cd090c79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.58, 0.26, 0.07, 0.09], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9c0199a2cd1108d_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9052bb24eff5a39f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.06, 0.19, 0.46, 0.18, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4562744848750fbf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.79, 0.14, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6cc706f07b50eb23_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f592711a691fa547_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.22, 0.71, 0.03], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_555c2537120e90ff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.18, 0.25, 0.39, 0.08, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce3f0d0ab4125653_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.15, 0.52, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d383dcd0814b7642_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.46, 0.36, 0.09, 0.09], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b89ad6a04f88023_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.33, 0.55, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b76d82a44b10201_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.64, 0.26, 0.01], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee0afb9e90934c11_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.26, 0.64, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_035f33bf819b2408_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.4, 0.15, 0.24, 0.04, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce18f8e67c266d79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.27, 0.59, 0.07], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92020e4efe4f1118_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.37, 0.6, 0.02], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6000742ec2c9fb04_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.21, 0.16, 0.39, 0.11, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f86866d2979e760e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.2, 0.02, 0.47], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cae845aabed7648_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.96, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933e7b4534733774_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5ee7b10138ad9b4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.68, 0.1, 0.21], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4cc27cfb8ed57f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.09, 0.36, 0.43, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc890de4307f65cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.32, 0.56, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93181c923f545bf0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.36, 0.55, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dde8333f9126638_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.51, 0.48, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b610cb03c2449cc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.36, 0.08], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8257a93d1597b331_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.62, 0.3, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ee7760733d7a5e6_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a1d28c0999ed486_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5a2161acb694e75_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.14, 0.17, 0.38, 0.13, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fbdd6cd272e07b7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.86, 0.1], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855817f0f262df92_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e721abe284679d12_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.11, 0.09], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90f8845d79f24995_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.44, 0.55, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cee2e5e9c7f165b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.14, 0.54, 0.22, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e52b22a99831d74_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.25, 0.48, 0.2], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e22164bab248b4e_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf20fe92ac4a9ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.13, 0.03, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6335985efefdf4ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.29, 0.07, 0.32, 0.14, 0.06], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04ec83834058a5c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.05, 0.39, 0.5, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ed52b0691fdd43e_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cec4633d1fc95ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6c9d1519ddf3c8d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.2, 0.38, 0.25, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25cadd571d3d41f6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.43, 0.13], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ddac1f8244e8757_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.25, 0.62, 0.04], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9f30ef8c32842e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7053a11502fa67ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.08, 0.62, 0.26], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0228e582e78dd335_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.75, 0.13, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0baca6dfb1c8ed56_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.85, 0.1], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc304ff26cacd3b6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.5, 0.17, 0.04, 0.29], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f1aabf902a63b59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.12, 0.16, 0.4, 0.18, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23543d9192e5f9c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72c3f40d6f20384d_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_873b5e2860432e39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.21, 0.71, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3cc45bb4c4a3111_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.77, 0.17, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9449ed737dae8b74_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e55ec09ee2560cf8_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_217f0fb19e9ab791_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.56, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39d3b654c286f635_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.04, 0.26, 0.56, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67bbceeaa11de3a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.73, 0.09, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cb9d90d0bcbdf1a_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59f5b0ccec05f32a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_945a4ee978383c65_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.3, 0.61, 0.05], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a48ac86858e07cc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.14, 0.08, 0.75], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_106fb08e5b5d172b_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_916b8cb2f81634d0_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fb36b0f54a8872d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.49, 0.48], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1de33dcf75a9ad38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.18, 0.18, 0.39, 0.14, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f566ae45ea7d0c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.04, 0.61, 0.09], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6891cb2fb12cb8c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.35, 0.47, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f94e1158f0d9a6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.45, 0.37, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f4df05aa49159b3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.49, 0.09], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89289d750feba09c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.25, 0.07, 0.64], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_816891706439634a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.26, 0.1, 0.39, 0.11, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a533e70855df6cac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.22, 0.09, 0.45], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fba3adb70e0e150_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18000b5f7aa7ec98_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.24, 0.22, 0.34, 0.1, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2edbfa54a492ca18_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.48, 0.39, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2242e6693ae6033e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.14, 0.43, 0.29], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da2e7592af73b7c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ab4cd676c0d9d4a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.22, 0.66, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dfcf196449abdf3_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5be74534ade65b35_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9153760f9f3ddc9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b08e91cbab6c2b13_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88f9c0b0eb9b5f0d_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d10dd1d1cd8d687_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd357df2882a04a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.15, 0.23, 0.43, 0.12, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2cca499fea7165f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.13, 0.53, 0.25, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52852103f9b467c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.1, 0.52, 0.32, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_788197a165507557_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c381678179387e84_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.22, 0.11, 0.67, 0.0], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87071155e73162a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.37, 0.49, 0.03], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3959267f2b71a2f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.26, 0.61, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7449dedabcd20767_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.4, 0.14, 0.27, 0.07, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03c241c77628042a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.05, 0.21, 0.45, 0.2, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0216b2aea4e8b6e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.9, 0.07, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_038509c3e84c2d2f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.59, 0.13], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3af0c9da946f9598_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.14, 0.71, 0.0], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37a423670bdccac0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.58, 0.31, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01ce05d760c3dab5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.32, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10992a7ac04bbf8f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.7, 0.04, 0.03, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4bfbf7f9f2686e4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.02, 0.81, 0.12], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_222c3451b5275f75_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d79bdaf2c936480_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c849f0135312bd28_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45731559be6b926d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.12, 0.68, 0.2], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6c6141a5aadb314_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.07, 0.63, 0.23], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b0e6754bff5807a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.58, 0.29, 0.11], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b93387ffcd5495e_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1888572dbbc2798_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.31, 0.56, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d48370daf7eeb9f_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0334399392b60f51_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.19, 0.74, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec6a659e9194cae3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.31, 0.09, 0.34, 0.11, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_852735939cefedfe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.62, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18627bd0b5793970_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.06, 0.89, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533c5b71ebce21b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.24, 0.63, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27845f38518fe091_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_644841e296a23901_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.29, 0.4], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24e6ffa6bdce6207_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.54, 0.43, 0.01], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89f4302bd069fd4f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.39, 0.04], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0505efdeba22e451_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7bfc494df59b5cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d3d3cb81ff91d34_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff1652fd8d7755ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.48, 0.37, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfacae87d46dcf2f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.19, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_776df851deb154e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.15, 0.6, 0.17], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4b6fbd9ee92d17b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.02, 0.04, 0.33, 0.49, 0.03], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c8317703b39efe9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.61, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9c0199a2cd1108d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.04, 0.37, 0.48], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d8a6aaa5cba476a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.62, 0.1, 0.27], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a49c31df3fce37a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.13, 0.48, 0.3, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc5e5ad4dca3ee2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.57, 0.31, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bce82894511a1662_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.67, 0.16, 0.15], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9738231a2e5e108e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.48, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a8b34d4c27413eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.49, 0.41, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fbdd6cd272e07b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.16, 0.18, 0.37, 0.16, 0.04], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_330df55032511b93_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.6, 0.32, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dabd819e65a9b59_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.49, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1064ba06334745fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.15, 0.56, 0.18, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c366bb8423e6d9c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfd182d9b88e6ab0_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e9e86be4b6f5a83_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d857804cdfc5c101_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96f33b72ef10a06c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.21, 0.76, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_578c75d233dd48c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_136c6e0264e153e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1890a59accfe573_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b13101125ef6e21_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.12, 0.73, 0.15], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_179afc8b13e9b3fd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.77, 0.04, 0.16, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_720ec3c4e38f81df_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.42, 0.12, 0.36], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4286f67cf12195e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.09, 0.51, 0.29, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01654f20b95b8a25_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.17, 0.39, 0.23, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbccb3ac89d43c7c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.79, 0.2, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02ef1aee0da9f8fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.16, 0.45, 0.21, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45dc989a35214dd9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.22, 0.55, 0.16, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca29bbab70b45150_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.88, 0.09, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f6482186d61d23a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.18, 0.57, 0.21, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fade8401009d8906_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82bfddea204734b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.14, 0.85, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c47fa69076819b9d_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f13aa2e50a2d2bbf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.7, 0.22, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cbf32fc7d1357ee9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.5, 0.26, 0.14, 0.1], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_873064f89c553200_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a2b56812de9a4a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.43, 0.44, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_483530cfc0e92840_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.29, 0.2], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c72fccb90787fa2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.21, 0.4, 0.26, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3643edb1f13a73f1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4aca142b334ba144_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.14, 0.3, 0.53], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41b0f7e09579be93_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df31e953f25a8415_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.06, 0.42, 0.43, 0.02], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2349a345668b42c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.29, 0.58, 0.03], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c2891f269fcd50e_n", "kind": "noul", "options": ["false", "true"], "target": [0.03, 0.97], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b89ad6a04f88023_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.51, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8ff7e68636b987f_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfacae87d46dcf2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e330674fcc5ed136_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.3, 0.53, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09ebf1016d26bc60_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.09, 0.77, 0.14], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0b1ee898b4a1379_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.65, 0.16, 0.18, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e387340aafa3246d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.26, 0.14, 0.42, 0.08, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ef28fe276a4fdb0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.23, 0.67, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7aa71f6e258e6304_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420b24b4451181b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe4f8dcc566aae52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.97, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d002aee1a007137a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.15, 0.64, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a23f465d0db74b8d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.79, 0.2, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffbb734c82a3d75c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.24, 0.19, 0.37, 0.08, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77cabe14e8c36052_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.58, 0.09, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_452723316aaed817_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.32, 0.67, 0.0], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4417a7bc1345cd6d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.65, 0.06], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cf3ffe988217525_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_222dec12a2dc24bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.63, 0.04, 0.23, 0.1], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e4b01640776a61c_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf3e93654c6ae39c_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39302dd0d428ed0b_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6793546761dcffd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.1, 0.23, 0.46, 0.12, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fec57df729f93aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a8d9df0c7ecc172_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fd119c481e316d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f13aa2e50a2d2bbf_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54d569144b0711b1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.83, 0.04, 0.12, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_885888423eb70e74_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3f61057825835aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.34, 0.57, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b89fb1e29b17e9a3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.12, 0.2, 0.39, 0.18, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_609379d67c1c7604_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d8b29c1ac8e7b08_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.03, 0.31, 0.58, 0.02], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f463741616540a1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.21, 0.13], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fc9620410df9d55_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b13914280653a16_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1fa8201759c41b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e99f1c0ed9282c4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.49, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ab6f8d21de8e109_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.63, 0.08], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93f28a2d641d2549_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63d5277473509627_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.4, 0.44, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc7b1148da5dd218_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a2e6ad0e07216e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f78da8915277b5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.65, 0.17, 0.16], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cfcf85c28948c68_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.22, 0.47, 0.22, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_707f6746805be952_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.2, 0.28, 0.04, 0.48], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dddfae548712705_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.41, 0.51, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8910c14f60798a2e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.11, 0.45, 0.4, 0.02], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ba5790ddedb323e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.42, 0.31, 0.13, 0.14], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e937637420bb33ff_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f5fc61d42cc823e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.33, 0.53, 0.1, 0.0], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d33dcbdb17e5180_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.1, 0.42, 0.39, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15324ff852a46665_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.14, 0.43, 0.36, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_038509c3e84c2d2f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.06, 0.68, 0.21], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2ee6eb59258f0f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.04, 0.15, 0.42, 0.28, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_adce541271349f46_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5dfd0c3a1d1267b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60b06e1f0ebb154f_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25d2386ceb990356_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37f8afca792ff0f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60e9c0bb28a3793c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8, 0.19, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_517bd82e24735363_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0fca341cc136a09_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.29, 0.56, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53e11686442fc94f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.1, 0.37, 0.44, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_589f7011d258c4f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.06, 0.64, 0.28], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_945233792b032c04_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14f794220205ae2d_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a14af272649ab6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54fdf332a225027f_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5776e09ffb9c0feb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.79, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eaa209b839a4a32a_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ce6e4fb5d3fd39b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.06, 0.3, 0.52, 0.11, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffbb734c82a3d75c_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49faad74a96daf20_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0505, 0.0303, 0.6869, 0.2323], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca2e18e0a25bede0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.16, 0.17, 0.46, 0.12, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0bfbf686ffaf14c7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.24, 0.06], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b3b8bc63e89be5c_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf5bf542f98895cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.05, 0.45, 0.48, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf3e93654c6ae39c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.29, 0.59, 0.11], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff28c8fedfc8892_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.38, 0.46, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d255094533ec5b16_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.21, 0.7, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f1609e8bc644fe7_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6a044dcf0377ef5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.19, 0.25], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d7992c70c9b9475_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0b21bfc4172c472_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.2, 0.55, 0.14], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_807841ad0cc18283_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd4c2d46092907cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.12, 0.16, 0.41, 0.17, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eb7d16210199b8c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.4, 0.47, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54fb03c319d86867_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.27, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b9b89e991e62f4c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.43, 0.1, 0.03, 0.44], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8e65e15a7ab9c35_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.66, 0.32, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e35e68cc62497da_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.1, 0.18, 0.48, 0.16, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec317dbe447ecba1_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d3a551c157d1cc6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757e864169e8f701_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.48, 0.45, 0.01], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df63b4ba3d4332d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.16, 0.5, 0.25, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc533a7576106b3c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.22, 0.64, 0.07], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a9b93247e7e7295_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18f06aa2a7033b90_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.0, 0.08, 0.81], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccedb3e74554e314_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.89, 0.07, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3c7a78d9e8bb81d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.19, 0.52, 0.19, 0.01], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa3bfc501dcd439a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.69, 0.12], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2307369a90ee683_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.3, 0.03], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1361a0d30b11bdc_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6d2d05172a1e447_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.43, 0.42, 0.01], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91255cdbd80c2f7d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.45, 0.15, 0.33], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b61d59616769564b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.14, 0.77, 0.09], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc2e564303995ea0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.27, 0.42, 0.21, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b21642dd21a2c5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.7, 0.19], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_529aad22c700fd6d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.05, 0.06, 0.34, 0.43, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21acd689e4f191da_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7384579268b43d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.17, 0.72, 0.06], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b5eff964580e606_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.06, 0.77, 0.11], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e4cf06d1de55e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd45373dc6ea685f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.19, 0.63, 0.09], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a75ef54655d246cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ab8a43409dd8c89_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a39aa2ed422e7089_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.16, 0.5, 0.25, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_327e697625eb54a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.11, 0.16, 0.42, 0.22, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cae5e8e891d19bda_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71b8342812e437d6_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1e62540d9d699f5_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5eadc21fb9e2e6a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c76c29fa97f8dfd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.01, 0.97], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0386211b077ff799_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.35, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25cadd571d3d41f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.18, 0.45, 0.25, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_520da7aa3f73c772_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.21, 0.53, 0.2, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bced317f1b3f6edb_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eef4538f87ee3734_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.85, 0.12, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12558ba660d576bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.17, 0.17, 0.41, 0.13, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a9761704d49df88_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.1, 0.38, 0.48, 0.02], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89c00a58a35cc70e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.48, 0.41, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1bae8ddaf9d153e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.51, 0.18, 0.29], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36c1b8f67ec394a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b235f6be31f14a76_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61c0d40d4c581382_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d047073db6dacfc7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.03, 0.82, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f17549934649ab44_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0101, 0.0303, 0.3434, 0.5455, 0.0202], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_317a2c18da54412d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.44, 0.34, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb232c9c0ac2b121_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.35, 0.55, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c26adcfb2d68073_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.02, 0.11, 0.56, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_decd5f546f09c6de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.11, 0.18, 0.47, 0.14, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6ff94256d3d1abd_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_feba41da31699030_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d4fa72aa5dc730c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.05, 0.54, 0.37], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f9fa3bc219ec138_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df768186ac175fb1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.32, 0.49, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_321487005fce5a72_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.11, 0.6, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9873a983b7c85a74_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.48, 0.1, 0.41], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97950778d067cc09_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.12, 0.63, 0.21, 0.01], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39d9f6a46c868969_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.06, 0.44, 0.43, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9186f39e644d0c75_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b1b0fa116a38301_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.18, 0.76, 0.05], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9b0e8e627a71d9c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.82, 0.15, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b81f9891030cf43_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4cfbe1e7693abbd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.33, 0.0, 0.02, 0.65], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec44f25c8cf60d93_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67c7fd74dbcff67e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.37, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d247073ff1bca4e0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.04, 0.26, 0.58, 0.04], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d92860331691a23_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31bd35cdb43705a1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.05, 0.63, 0.18], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3094dd51c925df54_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caa252b01f5074e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.26, 0.56, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f24c9fbd9a436fb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7630f97850e9cb8e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.5, 0.45, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea00f02b194aa3ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.03, 0.09, 0.4, 0.37, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df4853dcb77320ec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.53, 0.1], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f0ab5005e531c4a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.37, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0511140a3ffd75ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.41, 0.43, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c925c51d084a76a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17c9f3dd0cc7f028_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6a43b48e74e2fde_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.1, 0.75, 0.13], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc44430dd91c9c6a_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05c1bbd94a26ed38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.16, 0.72, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4538177ef9217703_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df768186ac175fb1_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44bc98bdfe416ee5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.35, 0.54, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e1fe486dae0188b_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b25e14c861f3e393_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aa4cdc8551e28d7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.21, 0.47, 0.11, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21159ca42479a861_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e0f550f12a0a006_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.11, 0.53, 0.27, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eaf4459ebf6a8843_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.83, 0.08, 0.08], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d8e390b021d0cad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.05, 0.44, 0.27], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a01b96f27bfc0b29_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.82, 0.13, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ab5847543a2f6b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_568edfd33eb3427e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.14, 0.51, 0.29, 0.01], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_822f019f3f6ab89f_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e91255d77650c15_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c1a7e63eb1d94c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.19, 0.49, 0.24, 0.01], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2ad5cb7a4702e42_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.19, 0.63, 0.0], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de0aea36329054c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.19, 0.41, 0.3, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b62885408dae16ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.1, 0.18, 0.45, 0.14, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22bdf7302348c29a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.52, 0.29, 0.16], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_661bab7f5225c8fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36378cfdb955dedb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.61, 0.06], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaadd861609bbd1b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.47, 0.45, 0.07], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad0a407a4bb2b1c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.25, 0.2, 0.32, 0.06, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcf267c86ee995aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_381ea21ab0ff3545_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.72, 0.27, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_779ac12d9306cafa_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cff6dd29686864bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c06c6af9f6a39ac6_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_954ea0f34ed94eda_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.1, 0.17, 0.4, 0.24, 0.04], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cced86e9ea336d82_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.64, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7ae8e9079367878_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.35, 0.13, 0.21], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d48806ad6db99c5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.27, 0.72, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fdac7ababff7397a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.31, 0.4, 0.1, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7ae565c1df8e900_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.06, 0.88, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99605f6131a71ec9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.16, 0.17, 0.4, 0.14, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a9b93247e7e7295_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.59, 0.39, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38559f2b11f4e318_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.13, 0.72, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c3597d810b8a9bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba2b7f5e0983a194_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.16, 0.64, 0.13], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_719882f82c0d1749_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.36, 0.52, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a75c4ef0ae189577_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.2, 0.49, 0.25, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1064ba06334745fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54d5fa1b307e9820_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.17, 0.21, 0.17], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d34c97eed7f654d_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_427892592507f7b6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.79, 0.19, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89289d750feba09c_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa7ff157e0adc5de_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.7, 0.27, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06e649b90ec18ece_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.32, 0.38, 0.22, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_550bec41935f4e32_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.52, 0.22, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee4559b273e3531f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.44, 0.04], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3597e69e3d0fcb37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.39, 0.45, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea2ead956484abce_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.54, 0.0, 0.03, 0.43], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8e19d2f422c039d_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a0a9c4b94bf5034_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67b0a9daa285081b_n", "kind": "noul", "options": ["false", "true"], "target": [0.2, 0.8], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11f1afa862ca794c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0505, 0.1414, 0.5556, 0.2121, 0.0101], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35b143a9f03a4cf6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.31, 0.66, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ae1e9a779088097_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.15, 0.67, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fa0c5a2f0520acc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.43, 0.37, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea5fbf11984ced73_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.38, 0.5, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0104129156e13a82_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.16, 0.67, 0.1], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fdd732093a8e4e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.07, 0.44, 0.4, 0.01], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b8b19879c12932c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.09, 0.43, 0.45, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db7d1d92d3a49ed1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.37, 0.38, 0.23], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e3c541ad8f4c6bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.16, 0.48, 0.33, 0.01], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b723de0b7be9fc7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.33, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c49845a9336ec5bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.66, 0.31], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_217f0fb19e9ab791_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac2e7a724d02fed8_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae10890ce4c5ed94_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.57, 0.32, 0.1], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eb14c7aae97de9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfaa15f085f75efb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.6, 0.08], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93967484cf776134_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.31, 0.27], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5be74534ade65b35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.45, 0.36, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1726f6daf1e8381_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.17, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a72005ffc92732e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.61, 0.03], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e222807b809d909_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.6, 0.18, 0.05, 0.17], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c381678179387e84_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.15, 0.17, 0.35, 0.16, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ea81dfdb02e9cd9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc76a9c4a0b2753d_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b558e7427c84d58a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.58, 0.34, 0.08], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d315f10ccdde57_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a542802f05a29c80_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c708c9f85e4a782_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_864536b5b7330482_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.18, 0.51, 0.18, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abea6d3f20a042c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.11, 0.14, 0.45, 0.24, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5afc55e0c8ecdf55_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.09, 0.83, 0.07], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55bb6351f9613d89_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_548c00fcec078b03_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.27, 0.21, 0.34, 0.1, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec7da8e37c0755b6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.07, 0.58, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb29893d55dccaeb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0404, 0.1313, 0.5455, 0.2323, 0.0101], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0127b64efe39e61_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28596c5a5e2dd012_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.37, 0.59], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22bdf7302348c29a_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da2e7592af73b7c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.13, 0.7, 0.11], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b36784575ce0133_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08fbec55903b1a53_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0404, 0.3434, 0.5657, 0.0202], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40836617b69dbae4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.07, 0.15, 0.37, 0.29, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958b058965c1b030_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.62, 0.13], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaa35ebf336be0b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f6482186d61d23a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.22, 0.12, 0.49], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba334f417b47b8f0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.44, 0.53, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eed868b0f0a5eb9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.42, 0.54, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_423f0dfd3bfc5933_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c57f86afbe9b0dd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.55, 0.35, 0.09], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c53d243a75d1d3d_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f0ab5005e531c4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d5d96be856e6e73_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.4, 0.09], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d6a191ab8d94287_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.29, 0.06, 0.29, 0.14, 0.12], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c7efdb32bc229a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.39, 0.37, 0.23, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbb3ff9c5571375d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.43, 0.49, 0.01], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eb20552768160c34_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.74, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb4aeb00f54597b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.06, 0.82, 0.12], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8858b20ce5f77ec7_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae50a32ece50fd7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_154f01755949dbd1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.0, 0.11, 0.75], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_100bbc41f412975b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.9, 0.08, 0.02], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6379967d2da0806_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39872950d2be951a_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4dede90509c06355_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.15, 0.39, 0.44, 0.02], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_410f5b8286dd1ebe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.29, 0.65, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd68627f81e01e73_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.28, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc5dee796b7a9d6a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.27, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_838f852d34e0be1a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.18, 0.47], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d025b48904a764b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.1, 0.41, 0.38, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_897cf18127c0ab61_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.13, 0.18, 0.46, 0.1, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc4537c692ddaff7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.64, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e40339b2c1ba42f_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6923c78ce5427ab_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_795285a6a5af3871_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.24, 0.17, 0.32, 0.12, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5a710bbf06ff00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.17, 0.46, 0.29, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaa35ebf336be0b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.1, 0.36, 0.5, 0.02], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_767c6dc95eeb6fcd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.51, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a69bf16c7172a5d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69745af02b44ba3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c11ca2dc14e6cc3d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.77, 0.22, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dedb18810907edf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.15, 0.55, 0.19], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67fff2a3947190dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.39, 0.55, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0c1b9e43a9a23b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3229945a7f8469e9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.39, 0.0], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd161c26e9cfeafe_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67182dd324055568_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.14, 0.34, 0.28, 0.08], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c420e469c6ac521_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c76677ade9a3f58f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.79, 0.13, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8851929f7865297a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.89, 0.06, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24aaf37feb7d3209_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.31, 0.08], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b057d38289de2ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.26, 0.63, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4743134b82746ac_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.64, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87bed3086254f7f6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.57, 0.06], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e2c10e7fdc467a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.05, 0.58, 0.27], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a97a1b2f37eb1b5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.12, 0.44, 0.32, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17936cebe37839cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b2fc6f878b8b9f5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.2, 0.61, 0.09], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6acb6fa1116ac9f9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5455, 0.404, 0.0505], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5a710bbf06ff00_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277212b08de14879_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad4ddb9c5f163744_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.48, 0.45, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4acfb628964955b0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.27, 0.57, 0.12, 0.0], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3f41afbbc8f09bc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.66, 0.09, 0.15, 0.1], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8abbeb68bc995b86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.3, 0.61, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69ef1e2607449942_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.2, 0.56, 0.21, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a13adf2b2684bb8c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.11, 0.42, 0.36, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_298eb9f451829a86_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.41, 0.13, 0.44], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a01b96f27bfc0b29_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.38, 0.45, 0.09, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a3bee018d02c6d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.1, 0.45, 0.38, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c8a587b88d26881_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.28, 0.2, 0.5], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33f7e0fb5b177a50_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.12, 0.49, 0.33, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_848d9d0dab8a56af_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.14, 0.44, 0.4, 0.01], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c6731465262a2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.08, 0.27, 0.57, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7237c523643d12d5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.56, 0.38, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c2ff97b6d31a85_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aece766e405639a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.53, 0.12, 0.32, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db17f02970a5ffce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.26, 0.41, 0.29, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b9515c4826bf534_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93ebf4b4dfb23aa8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.55, 0.25, 0.01], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2eb47c6115732e19_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.78, 0.16, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a58d4a7ca4c9d176_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.37, 0.43, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_609379d67c1c7604_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.05, 0.14, 0.58, 0.16, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26535e9f82f0557_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.3, 0.64, 0.03], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e18bd2267a4b5075_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.73, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e333d2f212cf3936_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.04, 0.26, 0.56, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f463741616540a1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.2, 0.45, 0.21, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e9169a7b5ecb796_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.34, 0.48, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e43c31528acbc00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.31, 0.59, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c443dee8ce24b032_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.46, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_846cf85124d7c59b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.18, 0.69, 0.11], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3184e6a0424a5d9f_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1e8479bba1fccac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.31, 0.56, 0.05], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2df20a4352a46d2c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.42, 0.19, 0.33], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44cf965567366988_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.03, 0.11, 0.4, 0.35, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0446bac7b6cdd768_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_374110c228179029_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.3, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad47ee9153c37244_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5f9e3f74bb45b17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.2, 0.2, 0.36, 0.11, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_659825ab9841cd1f_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc463a620cf009e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac2e7a724d02fed8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.22, 0.07, 0.36, 0.13, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4dede90509c06355_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.18, 0.69, 0.13], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2349a345668b42c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e79e0b50df176cae_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.38, 0.26, 0.14, 0.22], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec44f25c8cf60d93_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.09, 0.74, 0.12], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b44c56d2f74d0dd7_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcc6dd27f1235653_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.93, 0.04], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f6cb6c042d1a44c_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97494f677cbd71b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.29, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56f5463daad1bcab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.53, 0.34, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dd8ed7513e8c443_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.51, 0.38, 0.1], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6453b5e625ebd1d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942db47bbbd7495b_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84b3674e3acfab5c_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a8784d29d315a0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.36, 0.52, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45a25f691bac9246_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc60bb7eb2af6aff_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bfcc172cce29480_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75059ee885d50547_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.03, 0.94], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85511f46a92de3ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.23, 0.65, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2199ea34c82ddbfc_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dea890377f33e4ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.0, 0.97, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85cc0503dd474ac1_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a655c1751652bc71_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_088451fdb61194d9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.18, 0.61, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cd7770d8b076198_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caa64df60ad7a120_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.03, 0.62, 0.33], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0ab3d681769ef13_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3984b2c495c99fa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.43, 0.18], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9613876fa9a4bb0_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e1018185e1de50e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.24, 0.64, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d17d25d4cda37c0d_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b79b625ceeb976c8_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0ae71b367d3feca_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.42, 0.11, 0.46], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a5e762fa5c21ad9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.3, 0.17], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6647a8a6d9b5af4c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.83, 0.16, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ed5f86f56792730_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.28, 0.27, 0.42], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6891cb2fb12cb8c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_791c3e52311209ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.78, 0.03, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93814c9337a9b736_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.31, 0.54, 0.04], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc7b4f20b69ea5eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.29, 0.64, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0dc08387db13270c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.39, 0.49, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e8f99d6b0cbcc90_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.12, 0.16, 0.43, 0.18, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0874c321dd891bd6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.79, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0794cbd1b4a30f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.67, 0.3, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f46de790b36fa58b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.35, 0.56, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f49866d5c9129a42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.23, 0.65, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f83134c2ee8dab6_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2748b748dba9409_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.21, 0.42, 0.18, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4143ac2bd2d5ca78_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a16603bbadfb5b6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.24, 0.44, 0.15], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83544e0b95a7cc8e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.07, 0.31, 0.16], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7e5c90b588d237f_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edd61b957f1466d2_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8700d591eb5992f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.11, 0.44, 0.36, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4907d3374772f53c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.28, 0.02, 0.01, 0.69], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a186bea8aff6b6a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.13, 0.52, 0.23, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27461670616936af_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fec4778e9660d0a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.01, 0.95], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b5d5573f3135faa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.1, 0.6, 0.24], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_519819b6824bfe01_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.62, 0.25, 0.12], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c7a863d47ff9f12_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ae16604561b07d6_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23cd11d993aa249a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.42, 0.42, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76825080c8b87f48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.47, 0.46, 0.01], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_506a0c9f52174b3f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.46, 0.22], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_483530cfc0e92840_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5600d2a6da2baba_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bebd57a1b061ff9_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_784acf7120218783_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aa4cdc8551e28d7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.2, 0.02, 0.49], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8229c68b84f5a097_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.15, 0.45, 0.28, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0962459101d9e7c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.42, 0.48, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e56d42be7b345ed9_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8970eed6a1bafe0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.05, 0.42, 0.48, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17abb1ede5c0ddaa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.48, 0.38, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3ee5dee1bad7bd9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.27, 0.02], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c4444912d2b283a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6698e6c7d18ea30e_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2df5ce0eea7f7aec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.65, 0.13, 0.12, 0.1], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d5c6484b55e3ba3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.42, 0.46, 0.07, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e41a425416829a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.03, 0.15, 0.57, 0.16], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd23192908e658ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.22, 0.62, 0.14, 0.0], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec2927ab3f6f40f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd1650b3b3e51360_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d619521db9741089_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_601339f43d2072fb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.78, 0.17, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b18e67555da367b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.31, 0.09, 0.32, 0.06, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1fa8201759c41b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.27, 0.09], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97e27660e51feeec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.22, 0.48, 0.28, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec7da8e37c0755b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5561e740ebf0cd4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.11, 0.03, 0.14], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84fb23aa0100ef2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.03, 0.26, 0.59, 0.06], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a186bea8aff6b6a_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_754c9e8b6d9352c3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.73, 0.2, 0.06], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e570fa9aa32ad948_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.77, 0.15, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4eecb6fbf12d6e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.42, 0.44, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0d6226beaa81675_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.75, 0.05, 0.17], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1061dfbc751d1fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.05, 0.4, 0.45, 0.02], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72fe386981e43744_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.05, 0.9, 0.04], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2cca499fea7165f0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.62, 0.19, 0.15], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ef03dcf92f458c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e704a65f6f83cfec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.38, 0.01], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6793546761dcffd_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d1cb3dfc7166aa6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.45, 0.43, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6954279c23320ab6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.08, 0.22, 0.38], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd8bd91b2322284d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.04, 0.64, 0.16], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f08d8ac1dc28333_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.57, 0.37, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9cf9e4b41b6ce66_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.64, 0.15, 0.02, 0.19], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01654f20b95b8a25_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.43, 0.19, 0.23], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6c3022ddfa9750a_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9a8789150f5d774_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.32, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eed868b0f0a5eb9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.05, 0.92], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da00307b841ddd7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.36, 0.44, 0.04], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e121a7fd620c3bca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.19, 0.23, 0.36, 0.13, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75fbf17ed50a98e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.64, 0.03], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b62885408dae16ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46478b26415bf170_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.11, 0.58, 0.22, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3da18e91917ce4c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.36, 0.54, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a850d0a86d295d1d_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d6491d630e354c8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.26, 0.17], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4848f70f610b5371_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.49, 0.19], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aa3991616910d45_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.17, 0.29, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0962459101d9e7c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1353a18bc0141bd3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.22, 0.49, 0.1, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f550d0c687a97a79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.1, 0.43, 0.4, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91cf0f67fbb22184_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.78, 0.16, 0.06], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c00c0f6e22d6e00f_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b946cfc07f7090f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.27, 0.39, 0.24, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1217cc4e8c6381e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.28, 0.11, 0.03, 0.58], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13c9b9f944881c4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5100fac467897a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e061b301f3f14733_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.69, 0.24, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8a4188f07f4d3ca_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.6, 0.17, 0.02, 0.21], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_333992169e79c9df_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.47, 0.34, 0.18], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66ac27615660abda_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.48, 0.21, 0.26], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67fff2a3947190dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6bac7308f60c5a6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.2, 0.79, 0.0], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19b9f3224fc3992f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.46, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a5a9421a2ad5d9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.1, 0.41, 0.46, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17854a82cbf7ce39_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea970192e7f14d83_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bec620e56256871_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.95, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5799aa76d2d52fcf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.53, 0.07], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bec620e56256871_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.03, 0.05, 0.21, 0.48, 0.09], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_977f892687670541_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.48, 0.12, 0.31], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44bc98bdfe416ee5_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c4de750e43ec9cde_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.25, 0.62, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457b7ea1440e0296_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.46, 0.23, 0.07, 0.24], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5dc6b384a91f22a_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d4be9081a6cb9c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_effb9fe937e0e689_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f0deb9bd988c07a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.37, 0.61, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d3450964d628f83_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.51, 0.34, 0.12], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13d0b4d32cc778a2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.63, 0.06], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e94df78dcaf13dac_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05e965e825eb38e8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.21, 0.51, 0.15, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f088c691a359dbd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.35, 0.06, 0.21], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d3b3a97122124e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8b967f8aeed4929_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eff659e1a3972b0f_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d0432a7e80c144c_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2a205b636942a70_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed4b1a372acfcfe5_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1056fb002cbe3e62_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.05, 0.65, 0.27], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6e1f5d4ae235bd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfe5f723623953f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f27048f5d1ff451c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.48, 0.3, 0.09, 0.13], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e135faaab793be7f_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdaed520e703c97b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.01, 0.96], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02bca45a9df9833e_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b5b829bb4db5db4_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58b83624b3f43211_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_baa5756b3e67a276_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.19, 0.07, 0.45], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e91255d77650c15_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.3, 0.47, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86a2054841463def_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8d76212c20f55ff_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a2e6ad0e07216e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.11, 0.18, 0.41, 0.18, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_961f2b8ac03fcf2a_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebc6eb08cc13b72c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.62, 0.04, 0.01, 0.33], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dfa46782de59924_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.12, 0.48, 0.36], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c49ac7c6a3c733d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef939d6075dde301_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.17, 0.07], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a96fde324f2d8f87_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9dd70c40a315caa_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da1cad57f0682d87_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.29, 0.22, 0.36], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d37ce6a5c5e44fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.15, 0.7, 0.09], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f745051580a558be_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90fb8ac3fa895a3f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.42, 0.28, 0.28], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c4f8af6914355d83_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b55f23bf3fd4508a_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edce78236cdd2fed_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c88659e68e011d37_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.79, 0.12, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_30b83d5595603d9b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.27, 0.73, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b8a795515606e2a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.47, 0.5, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d53c5f1c36e0c9e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.59, 0.21, 0.18], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1401d87832234d3b_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5add67e3ac8bfeab_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.59, 0.11], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca32a3b923e4da7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.31, 0.59, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a40ec5dd180fd28_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.09, 0.45, 0.35, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e24c6c69d5d1de3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.25, 0.11, 0.39, 0.11, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_446f3dbbadc0cf1a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.12, 0.16, 0.34, 0.25, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b7e816f3e0e6ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.22, 0.74, 0.03], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_327e697625eb54a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.06, 0.93, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5bc880b1acd2086b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.77, 0.1, 0.12], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9060c353303d6444_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f3715b5e85501ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.12, 0.23], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5011ecb83f3444_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.42, 0.22, 0.26], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c11ca2dc14e6cc3d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.59, 0.24, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21fcf6e75dc00c2f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.4, 0.45, 0.14], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8a41bceddc18878_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.25, 0.69, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40a4b720aa7eca0a_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc76a9c4a0b2753d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.8, 0.13, 0.05], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_811d577aa5227c54_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_178fcabbbb427a4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.11, 0.46, 0.37, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d96279cfbdd1c689_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.47, 0.04, 0.35], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b144b32b8727d1a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37f72e91ddf9eddc_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31f9f37481d2d97a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.15, 0.39, 0.34, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c440e4bfdd26bf2c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.06, 0.93, 0.0], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de44ea1663f30b7e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.17, 0.73, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26535e9f82f0557_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.63, 0.28, 0.07], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f797452272707f42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.11, 0.63, 0.2], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_766aa1893edc931b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.63, 0.17], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a4a0a4e6c7e64e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2ed0f23ba28ac64_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05e81f95f90fa438_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.06, 0.52, 0.36, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c31fe80e71534ab2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.97, 0.0], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4a0387f351cf745_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.7, 0.07], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb8cff9477c0ab46_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_facd826ee4d3b772_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.67, 0.16, 0.03, 0.14], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8c6662a2840fb88_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_937788fab522a730_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.35, 0.04], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82711e257e51cccf_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5757f1b17f3c81_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.08, 0.6, 0.19], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bfb609893d01230_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.06, 0.72, 0.2], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54fdf332a225027f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.12, 0.25, 0.54, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7045e0903742867f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.8, 0.08, 0.1], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c76c29fa97f8dfd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.55, 0.32, 0.01], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6c664550416bde1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.02, 0.56, 0.33], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c302e9295a4a516e_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da1cad57f0682d87_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7d9ca5f042e8a9b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.0, 0.99, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f3715b5e85501ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d03a7eb0d82e64e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.38, 0.22, 0.39], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_125d5388a613cc6d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.12, 0.29, 0.53], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f7e4a8a3ed27a6c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.56, 0.3, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f36e3f3387d1bcc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.17, 0.49, 0.32, 0.01], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e0349440416b9d3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.27, 0.03, 0.52, 0.18], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d0f5c016f621589_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.65, 0.33, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aac55707642c142a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.19, 0.78, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c603c56425afce81_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1268d0b86551ff3c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.0, 0.09, 0.85], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_729cb0865cfc328c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.98, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcf3a0a5f2ce7e04_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c21709c7047ec6f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.4, 0.47, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6087a465590c140b_n", "kind": "noul", "options": ["false", "true"], "target": [0.17, 0.83], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35220e8e83d2c5cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0e7381c471c3c99_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56a9b45a7d9daa5f_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c0397515dae2b5b_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49109cc752fa4365_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.12, 0.61, 0.2, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24aaf37feb7d3209_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_668bc2cafea8dcd6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.78, 0.08, 0.13, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f960e297488897a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_658709dcd9250ee7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.44, 0.09, 0.23, 0.06, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f463741616540a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae350bfe85dd0516_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e07f4b429ed48587_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.17, 0.7, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec7da8e37c0755b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.15, 0.17, 0.35, 0.2, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_217f0fb19e9ab791_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.13, 0.45, 0.37, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33046db768a0b74e_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba5f09285e254a9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1afbe6ef6d34f17_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6453b5e625ebd1d5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.12, 0.12, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8232747b952b2e8e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.12, 0.41, 0.42, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4466c7a3fea88fbf_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958385dfbf30315c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea787f435ab126cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17c79ada3b6cf0df_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaec8b5eaa79ad4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.38, 0.5, 0.08, 0.0], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de606aa6ae67d732_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36b736b87b466f18_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.06, 0.91], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a96fde324f2d8f87_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.09, 0.16, 0.52, 0.17, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_578c75d233dd48c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.21, 0.0, 0.01, 0.78], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa5f2d01df0b2d14_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.59, 0.03], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4005eb7b441c70cc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65a29a7094c8f953_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edb8a8cf5c081e94_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f80d66a3be05d890_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04bb0a061b179071_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.36, 0.45, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e866fb21121eb12_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.24, 0.7, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0ba375513d5c1e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.19, 0.46, 0.28, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20bea058ac26bca9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.08, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0682e7fa5e948e0_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75816a44b8267847_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.68, 0.22, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df79e920cf5d75d8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.12, 0.88, 0.0], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0792432ad7f0fd8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.17, 0.27, 0.13, 0.29, 0.09, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44bc98bdfe416ee5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.4, 0.08], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e721abe284679d12_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.05, 0.4, 0.48, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_489b4b3596340e77_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.22, 0.56, 0.16, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9ef3836e42bf110_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.35, 0.54, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96c68a11d667d777_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c543ea17e1823f88_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.3, 0.17], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7be40e18def55d85_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.22, 0.58, 0.19], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d998f63507f7b47_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_632cd17dd8f30632_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.23, 0.58, 0.11, 0.01], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0c1b9e43a9a23b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.02, 0.64, 0.34], "state": "An agent plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d9716fa916e10e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0101, 0.0303, 0.1616, 0.6869, 0.0707], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_850f830e3f465847_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.18, 0.74, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17373deb521a8227_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.18, 0.36, 0.39, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a6e74bd345b9aa2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.87, 0.0], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c84b178eed9bb32_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c6542da9aae1377_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3a8b8b6d0cb4aa4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.04, 0.95], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f812f5ca76c85e4f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.79, 0.14, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cccfc9d4693db1fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc8e3a7d7552371e_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a67953a7b5447c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0139a653370db058_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d1b72dbe90212ef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.44, 0.47, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_024ace9bc5ba2e79_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55a647e60c7d6ae5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.36, 0.49, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b255d7a2a2b38490_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a16835eef616d23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.33, 0.42, 0.24], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f27048f5d1ff451c_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2c5ff337ba296fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.37, 0.47, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1274b5e329255a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21159ca42479a861_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.34, 0.17], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e01e2d70105cb419_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.15, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c925c51d084a76a1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.63, 0.02, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da9d31fdc7bb7e7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.4, 0.19, 0.26], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75405c9a48fe73ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdb7f65c07475cec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.4, 0.14, 0.35], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5ae31fa0847e941_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.31, 0.38], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_258b0b8ca5f32121_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b56d080ec1e3f63_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.17, 0.68, 0.14], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb169bccf2c60e39_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.52, 0.42, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73e4b79db552ef22_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.48, 0.46, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f23b236f960605c3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.51, 0.08], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92d2d03a7eb1d4f8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.08, 0.29, 0.56, 0.06], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b3edef5da93d2b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.33, 0.49, 0.1, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bde2f43ac382cbbc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.14, 0.55, 0.2, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_863aaf03c6236e8e_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60b48e3d20a50666_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1d23118b7ba63ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.2, 0.64, 0.1], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a145377de56e4d_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2efc4ee6111512b_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bafa7e660d147f48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.1, 0.36, 0.43, 0.08, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1659a620a5358cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff1652fd8d7755ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933e7b4534733774_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8, 0.15, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e303bb01e3307f3_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29271f3582cbf5b3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.3, 0.01], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dddfae548712705_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_736d4923cc9210b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ae16604561b07d6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.08, 0.33, 0.53, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f68e4963b6bca43f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.15, 0.43, 0.28, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9e08cb9d3dc50f2_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f93c42cd29312c_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2cca499fea7165f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24edde6920904b2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.2, 0.51, 0.25, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbbbe0432ff72e52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.31, 0.56, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc7b1148da5dd218_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.14, 0.18, 0.44, 0.14, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee0afb9e90934c11_n", "kind": "noul", "options": ["false", "true"], "target": [0.16, 0.84], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a58d4a7ca4c9d176_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.33, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8bfa2d9adab1783_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.17, 0.42, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c16ed93f98e9c0cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.48, 0.33, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28596c5a5e2dd012_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95702db19e5b02fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4f8a800dc49aca1_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9b0e8e627a71d9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "A proposed workflow includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a23f465d0db74b8d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.55, 0.32, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c7efdb32bc229a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.48, 0.34, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f725e4d9636f3c0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.31, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea49a7fc9f04c649_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.02, 0.43, 0.51], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cf925dc454c1e30_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0dca1c074659eed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.37, 0.12, 0.31, 0.08, 0.04], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5600d2a6da2baba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.02, 0.04, 0.18, 0.52, 0.13], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fbfb624f7c1325c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24090441f34b8ae1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.07, 0.78, 0.13], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85259c65fde0a38d_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2efc4ee6111512b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.6, 0.31, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dba04db8b39179a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.01, 0.59, 0.31], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea49a7fc9f04c649_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.11, 0.44, 0.35, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4afa5095241da70_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.27, 0.2], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5be74534ade65b35_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.27, 0.15], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3929c4c582d3edc_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5549ccef0870e06b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.27, 0.6, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d388da3704f721_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0b2c820238e7587_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.26, 0.16, 0.27, 0.11, 0.04], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcc1c231722a9fd7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.43, 0.03, 0.33, 0.21], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_634cf83132e0a41b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.06, 0.26, 0.19], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e06a74c879aa64_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f81a22a24cd078f_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f62571168dd90923_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.3, 0.51, 0.15, 0.0], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a81d184401dc47c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ad7cd62b20d0dc4_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e1f7a2109ed89c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.43, 0.41, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1c4766d6a22c512_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dc392d28da02276_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.62, 0.06, 0.31], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_500b156b9f659e7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.17, 0.5, 0.3, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_168f90213ca83bab_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d466ad6788286eb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.16, 0.83, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_119a4781b9e3fa2b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.36, 0.39, 0.05, 0.2], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92f4fe645926f762_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.07, 0.65, 0.22], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1e3611a90fa1ecd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.47, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee53640af2d62757_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c1e123c10a72a48_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70b3fc788b4930ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.32, 0.5, 0.04], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20a969f6e9f1e409_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0101, 0.2828, 0.6768, 0.0303], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f4253883d3e6d55_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0404, 0.0, 0.0202, 0.9394], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b17db984026b6c04_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.38, 0.47, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a542802f05a29c80_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.73, 0.03, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_067018c50b7b46cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.23, 0.72, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_379442308dfedbd7_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08dced6b7d058edc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.35, 0.63, 0.01], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f0c06f473f05483_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62409e747b7d6f5e_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c3fd4760f62c226_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.26, 0.59, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc3abd12576d7219_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eee361e9a9332d0b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.14, 0.55, 0.28, 0.01], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8cefde14f961d9e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.25, 0.64, 0.05], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66ac037008bc3439_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.54, 0.36, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4dede90509c06355_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27361b65af035d09_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c828ca053ec974f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.2, 0.21, 0.44, 0.09, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2a2d83a49628ee2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.11, 0.26, 0.45, 0.07], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d327b5dbbd3805a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8a27da955a31497_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.31, 0.53, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_751d7465018b3bf5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.13, 0.59, 0.23, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_864536b5b7330482_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a69bf16c7172a5d9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.0, 0.05, 0.88], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_feba41da31699030_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.06, 0.71, 0.22], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3841fe515844e2b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e4ddf6b03025a86_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.3, 0.06], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc97c5ea9f2e2409_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.33, 0.34, 0.32], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f468b1e88e881963_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5891a2ced68dec9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.25, 0.16, 0.37, 0.08, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e18bd2267a4b5075_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ccc5be05c7b18d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.38, 0.54, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3da706bde7c5d4f1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.51, 0.3, 0.17], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4562744848750fbf_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ac6e3cf729a318c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.14, 0.47, 0.24, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ab0765156c68db9_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d79e03b47aa640_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.36, 0.44, 0.04], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a1d403c0f5d2ae7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.57, 0.31, 0.11], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d00f6a757531e258_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.42, 0.51, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f844dbc3827c0889_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c93c08e5ff47243_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.23, 0.48, 0.18, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75fbf17ed50a98e_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6cc706f07b50eb23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.79, 0.1], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69210ebcc6ccfdc4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.17, 0.69, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97950778d067cc09_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9f1e23f7cc8ba44_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.52, 0.27, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c80035513f429bbd_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d95fb11bbbfaace_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.14, 0.53, 0.21, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95702db19e5b02fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.15, 0.39, 0.39, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbf218dbc273ad2a_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75059ee885d50547_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_596e0baf008eb976_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.27, 0.72, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7449dedabcd20767_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29d3c5e68fee195d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.18, 0.45, 0.29, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7214566c4380449_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.3, 0.7, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8dad925a0816bf17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.08, 0.4, 0.41, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a25336df8d6d89ff_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b144b32b8727d1a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.37, 0.53, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57a6cfc3de4e31ff_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.71, 0.12, 0.13], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4495c0424baafe5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.45, 0.51], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9f30ef8c32842e1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.03, 0.95, 0.0], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bbf739a03afcc0e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.1, 0.2, 0.43, 0.17, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1361a0d30b11bdc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.62, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e9be74f53c3e129_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24f1643ba878b640_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.09, 0.01, 0.88], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d96279cfbdd1c689_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.19, 0.49, 0.12, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abf03e6312779830_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39dbb8fce4ff9aea_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.34, 0.46, 0.16, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b01d579bfb13a303_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ef03dcf92f458c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.39, 0.48, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65a29a7094c8f953_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.1, 0.16, 0.36, 0.22, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a4427b786583342_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_063ff4acd09b4cfb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.16, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b432480ee5d1d219_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.44, 0.31, 0.22], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45d65f9f52d26978_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.57, 0.29, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db76fee0b259e6cd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.09, 0.51], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccfbdd79ecaf5795_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.5, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8084730dedaf4734_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.18, 0.41, 0.33, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e42e7fc43ad716c1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.59, 0.28, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09cbc322465869dd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.72, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18ce66bc6c378fcb_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fffb1a98c8f69ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.23, 0.54, 0.19, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a439926d35bddc7b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.09, 0.65, 0.2], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6648b26a85a46e7a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.11, 0.2, 0.46, 0.15, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05863c271559edbd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.11, 0.36, 0.41, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1c6099a22aa3e5d_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ce01c961ebdf743_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.52, 0.06, 0.38], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_facf915526833856_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_586530ef2ad24e6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0104a22dc61ed6a0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.31, 0.66, 0.02], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b855a855ec78f96a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.68, 0.04, 0.25], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56c4064674cedd40_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.76, 0.23, 0.01], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35b143a9f03a4cf6_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f411c0500168614_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.03, 0.09, 0.44, 0.38], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b42523d1ff91c900_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e97e2ad8dfc9749_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.83, 0.04, 0.03, 0.1], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_154f01755949dbd1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.29, 0.59, 0.03], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68c4bb1c1eaa6238_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_802cd2a26e91ba13_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b99e799868e6991_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e8cb07fbef09878_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2a2d83a49628ee2_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c84b178eed9bb32_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50abb6b0785e2ef1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.51, 0.29, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b238f8e8ab1f8f36_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.7, 0.19, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69ce1c7f0f294e09_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.26, 0.63, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c97524cf31e6d41_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cbf32fc7d1357ee9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.44, 0.45, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db76fee0b259e6cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f4187afc6f3ed73_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.36, 0.51, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f0afdd191626b26_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1414, 0.202, 0.2222, 0.3434, 0.0808, 0.0101], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6379967d2da0806_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.06, 0.73, 0.11], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75144fa60716d0e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02c9eef1af87fcc3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.08, 0.24, 0.49, 0.16, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277361bd393da474_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8229c68b84f5a097_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.92, 0.05, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d751ac598396abb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.16, 0.19, 0.44, 0.14, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ee10e94443f32b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.24, 0.21, 0.35, 0.08, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8462685895c549ec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.59, 0.23, 0.14], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97494f677cbd71b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.09, 0.55, 0.25, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ce3b9d955a7440_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.68, 0.22, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16e60230c0b8d60e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.1, 0.23, 0.5, 0.14, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e684df25821a7ae5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.71, 0.09], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_028039e1e25fcd15_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d48d9744aede6704_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bf4cef3db98ebdb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.56, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80410555622e2d3d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.66, 0.31, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41d9d6d888d9a6e7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.9, 0.08, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b66f015a29a340f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.17, 0.49, 0.23, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18ce66bc6c378fcb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.75, 0.23, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a36bb7a7287eeb_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_155492440990c9f0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.49, 0.12, 0.36], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_341846233cc4ae0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.34, 0.47, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7aa71f6e258e6304_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.12, 0.14, 0.44, 0.18, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_623cf5d202140573_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.16, 0.79, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f244176d64e3de_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15dcbcf7841c1b24_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.57, 0.27, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f0453b14234a47d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.19, 0.51, 0.21, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_233938d0d6441d2a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.06, 0.43, 0.41, 0.01], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7b11e3201c354e5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.12, 0.17, 0.36, 0.21, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89289d750feba09c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.46, 0.44, 0.02], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32b8f50877956722_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecd7cc7c64db8def_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.78, 0.07, 0.09, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2c6a9998d996b58_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.49, 0.23, 0.25], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd9e3f7b60f05a18_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.25, 0.68, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c63a9035bb816cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6153ab08913c764e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.81, 0.16, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3abd9133fc74e47a_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd4f16a66c178d22_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.33, 0.12, 0.54, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d07d5b782afb692_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.04, 0.36, 0.5, 0.03], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1955980e8928d16_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.31, 0.04, 0.34], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c1bd6823668fbb2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.36, 0.53, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_704f2150310ff950_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.08, 0.21, 0.5, 0.17, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6459d294028ee5c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.88, 0.09], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5243097fe9e634d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.29, 0.67, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23abde15345cdf9a_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_116aa6f2faac581e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.2, 0.06], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a05414d8252f8ae6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.31, 0.14, 0.46], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8fc72382c3f8da4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.17, 0.66, 0.09], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_961f2b8ac03fcf2a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.23, 0.61, 0.07], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01355a1bbe3ed442_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.3, 0.16], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a59d3fef0d1a6b33_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.19, 0.68, 0.06], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5945438dde376d91_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72b2f1099320a361_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67b0a9daa285081b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.8, 0.09, 0.09], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39fdeb6d5212dcd9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.51, 0.27, 0.19], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db44fb6238b20775_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.05, 0.28, 0.55, 0.04], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bbdf0505e4a71a1e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.81, 0.15, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afcc6160291fb4b8_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69b25b785070925a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2d6d766ec7ef188_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.23, 0.65, 0.06], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a7d9cd7eb64c642_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59c41fc94cfd5832_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d3450964d628f83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.08, 0.19, 0.43, 0.17, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_779dec894044c4a4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.73, 0.24, 0.03], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc96f4f36fd692b3_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fcf46e4e0ae9ba7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.28, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3b5418192edda8a_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e343315796609820_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8e26ce4e14eb363_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.01, 0.9, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_480095c0435e9bf2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.02, 0.94, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecac8bd4643bdc06_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.08, 0.5, 0.33, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4963decf96f76b75_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.31, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d95fb11bbbfaace_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.41, 0.1, 0.01, 0.48], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11d6ac10253a0dd9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.31, 0.6, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9934de50556b441_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.36, 0.52, 0.06], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_111f2f1ce23ab7c5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.29, 0.69, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40532f286c66359f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.14, 0.34, 0.38, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50b71c9cabc2f712_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.1, 0.37, 0.4, 0.05], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e1d13d0afdf691_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.36, 0.57, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5bc880b1acd2086b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.18, 0.51, 0.21, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f728f2b26b7fc214_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.47, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e54371ca4e877a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fe247362fe3b19c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.03, 0.28, 0.58, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54130c1ed1ed902b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.07, 0.18, 0.39, 0.21, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4cc27cfb8ed57f2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.04, 0.65, 0.24], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a609bc291f7c944_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee55b6d73875ff61_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83aee86b399ea3bc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.74, 0.17, 0.08], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f621483a700a1399_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74b53dbd2b1be6f1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.23, 0.49, 0.21, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e6f3adfea4b921_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.01, 0.19, 0.67, 0.11], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76504429f57f0da7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.26, 0.64, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_506aacd399e08900_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.85, 0.13, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_021b478b27fbe3ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.51, 0.34, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a7da70dbffd3501_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.22, 0.36, 0.18, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d2415f9c381d71d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.19, 0.64, 0.09], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e80bfd1d34c54e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2d6d766ec7ef188_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dc73f25b24fccf3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.47, 0.29, 0.02], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_339bf2f73c6e149e_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a558efff6c01c159_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.28, 0.48, 0.2, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d58df6bdbc8f988b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.83, 0.14, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e135faaab793be7f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.64, 0.31, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd693af83ab957e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.27, 0.13, 0.33, 0.09, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c2543e8f1dfc43_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.02, 0.0, 0.91], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a06245b6ec7b00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.07, 0.38, 0.52, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6003a4191b6cd42f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.5, 0.46, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c604e6dc890685fa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.3, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_783b3030434759c4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.97, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eeff4bf56ae110fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.17, 0.5, 0.26, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_760586154e17d289_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.4, 0.49, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3cb5d98b1cd02f9_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38ea826701498ac6_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_024ace9bc5ba2e79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.67, 0.2, 0.01], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aefbdd151872be0a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.34, 0.55, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02bca45a9df9833e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.02, 0.78, 0.0], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855817f0f262df92_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.12, 0.18, 0.38, 0.15, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fec4778e9660d0a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0707, 0.1616, 0.5556, 0.1818, 0.0101], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de146d67ec369d6c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.06, 0.01, 0.55], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42e7d70f881666e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.17, 0.2, 0.36, 0.11, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_787952bb1428591c_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d83fe27cb45f5f55_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.24, 0.65, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd29f42ef59295ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.05, 0.29, 0.63, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_021b478b27fbe3ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccfbdd79ecaf5795_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d48806ad6db99c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7484b95f2950c9f_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6216ca71d1f2dd5_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0113d0323359c538_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0303, 0.5657, 0.1414, 0.2626], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3ee5dee1bad7bd9_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_773018795f0555c9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.01, 0.07, 0.91], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3f400661d9b30f72_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31d31d3b5532172f_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ac6f9c5a9b9a09_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.79, 0.13, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_110b19e40ad41be9_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0e67cbfd369cf9f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.78, 0.07, 0.11, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a9d2209b76b8fe1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.07, 0.68, 0.2], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d5d96be856e6e73_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.08, 0.33, 0.52, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55aede0116108171_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.49, 0.19], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4619a29ab6084ab0_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e2bea7dd5b49f11_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.49, 0.38, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24478f5c28b98017_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.28, 0.69, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37f8afca792ff0f4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.22, 0.69, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4fcd1f6c69d31e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.04, 0.31, 0.53, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_858aa1a2838e1f38_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.5, 0.25, 0.22], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19cc9ce0d60170cb_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5dadb961d148c1b_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5820ca6ea318f88f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.26, 0.7, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2878faf404c1885_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.06, 0.28, 0.5, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6bca7419a25f7c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97950778d067cc09_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.22, 0.13], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cd7770d8b076198_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.01, 0.02, 0.83], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38f129a707ce886c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.16, 0.45, 0.26, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be685001890a23e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab004cf658bb0209_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0be927482a82c71_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.08, 0.25, 0.37, 0.25, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2ad5cb7a4702e42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.03, 0.07, 0.31, 0.44, 0.05], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f470afbe78692227_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.12, 0.68, 0.13], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1855d2e3a8897bf3_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f844dbc3827c0889_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_100bbc41f412975b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.6, 0.28, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cff6dd29686864bd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.12, 0.07], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bf4cef3db98ebdb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.25, 0.64, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b9cbfbb02b29a85_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.03, 0.26, 0.48], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cee5559055499396_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.23, 0.51, 0.22, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcbfc548f1f66d1e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.49, 0.42, 0.07], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ba37c74bed37a0d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.09, 0.39, 0.46, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc2fc02338f02fd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af98a9d512807255_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.04, 0.48, 0.46, 0.01], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6594a21bf8de74dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c2b55696ceaf618_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.1, 0.53, 0.28, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a381ea4de8b66554_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.0, 0.01, 0.22, 0.72, 0.02], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f4c42ddab15b25c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.87, 0.1, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef7879ac3e394e08_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.27, 0.3, 0.13, 0.3], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebee906dd88dc79c_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d94f6000e5bfb82_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_097b94e33ccc4a83_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_791ba2bd4d6dc501_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd3fd98514e3af96_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0baca6dfb1c8ed56_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.31, 0.18, 0.3, 0.05, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_721865a38745e4f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b93c73174df2fcb0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.15, 0.34, 0.45, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f9fa3bc219ec138_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.56, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c23af1d2520f75b8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.36, 0.48, 0.06], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e38911b976da5cf6_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d59f2f3bff1cafc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.21, 0.7, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4064e57265f37f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.1, 0.57, 0.22, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958da2319d0cdd38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.09, 0.56, 0.26, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_506a0c9f52174b3f_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3c6c082b6b09a7e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3, 0.05, 0.18, 0.47], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d1b72dbe90212ef_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b979252c26119420_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.28, 0.56, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d49d90ec1f220d1a_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2145178ee446a7f3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.1, 0.44, 0.28, 0.05], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ceb6527e146a8d98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.1, 0.84, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0007143882c4341_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.28, 0.5, 0.08], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48d8218c00f450a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.16, 0.51, 0.29, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03093030590dd0ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dde06d015c82765a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.15, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bd6c5317d2d45d9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.46, 0.11], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf55c73888d068ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0422fe0c4a2a954e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.03, 0.73, 0.04], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbe2cb93d6ec0290_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.65, 0.02], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17373deb521a8227_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_188401eb16308744_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a63899ffbc214fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b2f47be80766e5c_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90d6f7529640043c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.06, 0.86, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba24eb7453ad25eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f48d5bc7d205c154_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.68, 0.31, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_924ce83fe8655c2a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.13, 0.18, 0.3, 0.2, 0.08], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be4c0d9b40bef977_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.57, 0.39, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_133d70a53cc0117a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.73, 0.21, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_512cc169d3477acb_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88313f9c874fa8b6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.39, 0.01], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e78594bfb404a804_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.49, 0.33, 0.12], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c309cbd215074a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.22, 0.23, 0.36, 0.13, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_155614f8d77783c9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.91, 0.02, 0.06], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4099269cc5fa946_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.25, 0.19, 0.3], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_864505e3705ebc96_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.39, 0.52, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8baa60c91ac64306_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca3d7179e438f446_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.25, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b21b3c3bbeb5bb9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.05, 0.15, 0.4, 0.27, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90915e2a9c79ca9e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.05, 0.82, 0.12], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_300909b2ba845114_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.5, 0.26, 0.14, 0.1], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62a423b47e1cb7c7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.24, 0.06], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9915302d889050d_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93bc2f37bc45a28d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.36, 0.48, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1f0e06014104a45_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.18, 0.45, 0.24, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_119cd2f84a07011d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.44, 0.43, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97b1a72f226f8add_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.7, 0.1], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_825f1de029b6647e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.08, 0.14, 0.41, 0.15, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ca7671b35a6c416_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.58, 0.23, 0.07, 0.12], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bc0432619c91ab6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.69, 0.09, 0.2], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f960e297488897a6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.46, 0.38, 0.15], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b820ff4de9246dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.2, 0.68, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ed4e5dce24d7652_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5fd6903727bb14a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.64, 0.32], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d620400ba389a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.44, 0.37, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3313e203b53cdcfe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.15, 0.83, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1105cd2e76c98f48_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8, 0.16, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_299001dabd376d38_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56e340df1acc027c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.16, 0.75, 0.09], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92ecefa8b68abc93_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.23, 0.4, 0.35], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eb33ee004eaddc34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.1, 0.41, 0.46, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7b11e3201c354e5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.04, 0.53, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45e183baa3a202b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_500e8326fb40eb56_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.34, 0.05, 0.24, 0.13, 0.11], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc854bc563a88d90_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.11, 0.47, 0.28, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97dc77444b285ff0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0707, 0.1717, 0.5354, 0.1616, 0.0101], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d751ac598396abb_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c604e6dc890685fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.06, 0.52, 0.35, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e955c40d91d5857c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.11, 0.79, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc4a322fee89d6f7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.02, 0.7, 0.18], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47d6ee183a553824_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.33, 0.22, 0.41], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebabd95fa4217fc1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.04, 0.11, 0.45, 0.29, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d2a0010ae77930e_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5804b1ea89d925da_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.37, 0.41, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ac47a72328dc496_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.13, 0.74, 0.11], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_302ecbf772609d35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.15, 0.3, 0.48, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5118a20b3575ac4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.91, 0.02, 0.06], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9388b43b4d56d5f_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebee906dd88dc79c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.2, 0.54, 0.19], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40c7fcd15d0b98b3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.71, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6555216168a7a868_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.13, 0.46, 0.28, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b7e816f3e0e6ac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.24, 0.02, 0.72], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f45ac83d4a02031_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.45, 0.54, 0.0], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc628b56a69e612e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.21, 0.72, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eaf4459ebf6a8843_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.15, 0.46, 0.31, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a9d2209b76b8fe1_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2517c7bfde56b254_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.02, 0.47, 0.19], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_446546c30accb6bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cce7d2fca842f3cf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.77, 0.17, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69d964d0da328e78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.45, 0.42, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27cc1bb79b15da17_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.54, 0.4, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3dcf3fb0292189d_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7214566c4380449_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ee4c8c8cd008b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86210a6b8ffc46a7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.63, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11e8b6c8c6ad6012_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d13601f00d11c09_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.1, 0.77, 0.1], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9bffd8b4b4c318d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.05, 0.11, 0.45, 0.28, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_632cd17dd8f30632_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.73, 0.21, 0.06], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6bac7308f60c5a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.33, 0.15, 0.31, 0.07, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b855a855ec78f96a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.36, 0.55, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2cb7aa54e526b1b9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.47, 0.24], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d6f18318826c74c_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da1cad57f0682d87_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.28, 0.13, 0.35, 0.12, 0.04], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf5bf542f98895cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21f500b8b11a04d_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8dec5643dba879bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.32, 0.54, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cebe7973f99e3acb_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6016b048f187618_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95c1a1cb91750dfa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ac6e3cf729a318c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f40eaca60df888e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2605f16e337f331_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.15, 0.27, 0.44, 0.1, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91cf0f67fbb22184_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.15, 0.57, 0.21, 0.01], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b946f6faa72af7f_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fbd5ea0628b7444_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbc2ff581eb4b2e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.51, 0.37, 0.01], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d039f06efb7ba4b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.14, 0.51, 0.19, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5aab4ab386f4a1a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f353d1ffafaa60ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.19, 0.49, 0.25, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c700f379d568ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.05, 0.9, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d84206d71af939_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.38, 0.13, 0.28, 0.08, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e4210e866273a7d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.77, 0.08, 0.13, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c54dc84ffa7e8e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.12, 0.51, 0.26, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_672b283dfebaf19b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.15, 0.83, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aae0d29dcf2f08ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.22, 0.41, 0.11, 0.26], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb6cf4e29ec47288_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bc87874aaaa2a25_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.18, 0.5, 0.27, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b98afd8c74471948_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a948ed37dd97540b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.28, 0.67, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01ce05d760c3dab5_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eafcfcf2d5d7db78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.26, 0.53, 0.17, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664137f0f040ed34_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.62, 0.2, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1893c5ae8a303aa8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.17, 0.6, 0.18, 0.01], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dec5aa340555fe3f_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97ffbc1ddf4856ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2dd4ebb8f240660_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.18, 0.77, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcead17893dd7d0f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.11, 0.87, 0.0], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a47e0b6a296bccbe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.34, 0.09, 0.35, 0.08, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5958de6facfddd23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.36, 0.5, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49ebbcfc01d4a841_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d002aee1a007137a_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_802cd2a26e91ba13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.22, 0.66, 0.06], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7b903b6266888c7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.2, 0.46, 0.07, 0.27], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d472f467babd2c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.23, 0.77], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd1650b3b3e51360_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.32, 0.19, 0.05, 0.44], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aec8bd310383019d_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5212930c516f071d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.28, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05c1bbd94a26ed38_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dde758f62343edd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.33, 0.54, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46234e5c609bfbfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.21, 0.68, 0.07], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c17837781fd0663_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.44, 0.11], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a1d8e68b8596943_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.02, 0.87, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff20c7792692f2b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.41, 0.38, 0.07], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d1d48ab06b690cb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.39, 0.58, 0.01], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2eb47c6115732e19_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.27, 0.63, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_78b61e44ca22dd07_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c44f2f63cddcc8d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.17, 0.2, 0.37, 0.13, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a10bf0f4c787697_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_713fdde29303229a_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_175128c07b0e1b97_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.22, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9333317779ec4a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.3, 0.67, 0.02], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10ad33721f8bf53c_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe699100100827f9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.37, 0.04, 0.56, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7939721339970e4c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.14, 0.23, 0.46, 0.1, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f89c97182a70660_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.04, 0.2, 0.62, 0.07], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c80d28fe994323fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.73, 0.12, 0.14], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e3325b1fc0d35e8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.42, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7442c0fe0f2a543f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.44, 0.12, 0.27, 0.07, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e2c10e7fdc467a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_361dce359ea41dfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.11, 0.44, 0.36, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cef36f1b6365d77f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.1, 0.41, 0.42, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d619521db9741089_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.14, 0.19, 0.44, 0.14, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02dcdc11955fd5a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.12, 0.53, 0.29], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1315fb4846c685f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.28, 0.5, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96ec3a4040407cf5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.68, 0.09, 0.19], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c88b03890da55bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.38, 0.14, 0.28, 0.03, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b75b8cdbdc3be5f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.07, 0.14, 0.78], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15213e83e77648c6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.03, 0.7, 0.03], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2677a0192ebb72d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.18, 0.67, 0.08], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd3078beedb424ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea4ea17d3b05666d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.83, 0.09, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e563cdff265d1b3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.17, 0.45, 0.26, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fc68c1b88585743_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.28, 0.15, 0.35, 0.08, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40ef1e728335f865_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7b903b6266888c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24cf09a2c39c5017_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5eae093b10c6d286_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.08, 0.22, 0.45, 0.19, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c2de955856f4611_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.67, 0.24, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72c1790e619df6be_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.17, 0.15, 0.17, 0.42, 0.08, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7222b0ea154d4d3c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.26, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b6d4346390bbc10_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.28, 0.65, 0.03], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2c6a9998d996b58_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87e70297987fc237_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.48, 0.14, 0.21, 0.17], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1816c49e6443215a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.13, 0.44, 0.31, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a96d3994cf8175b1_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e7ccd7ed60234a0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.56, 0.23, 0.06, 0.15], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c061b9c3430adba3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.19, 0.42, 0.27, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2145178ee446a7f3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.45, 0.23, 0.04, 0.28], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93f28a2d641d2549_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.36, 0.35, 0.11], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ce3022e17dfde6e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.67, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf6ce920c3ad4c11_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d3a551c157d1cc6_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5dd56afa77554dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d88b973563b960f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.57, 0.21, 0.18], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0df22034803cf761_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.27, 0.63, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d734daba35e30d51_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c65c03fc3bfbca6_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae10890ce4c5ed94_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533f18762e88cc22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.16, 0.76, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16e60230c0b8d60e_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d235e54adbbbb6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.61, 0.35, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dee3ad473e590677_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.51, 0.04], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97b1a72f226f8add_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a35edc9ef07baa_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_514726ecff9f2185_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.54, 0.33, 0.11], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed3ce46a025d62e5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.9, 0.02, 0.04, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6bac7308f60c5a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d0d683794af532d_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b03835b56b94d32_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.04, 0.61, 0.23], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c2b55696ceaf618_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3377480ece924f4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.03, 0.32, 0.56, 0.02], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b476103258e042c2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.26, 0.13], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecbc1c68f3407614_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.24, 0.32, 0.05, 0.39], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23d3ff0e00f463c2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.0, 0.01, 0.9], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e62d031a17af2be9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.19, 0.68, 0.1], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_671ba4dae058fbf9_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fec57df729f93aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.13, 0.34, 0.39, 0.11, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b3b8bc63e89be5c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.52, 0.17, 0.29], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48d8218c00f450a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f87e544e6935872_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae50a32ece50fd7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.24, 0.62, 0.07], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dba04db8b39179a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.12, 0.47, 0.33, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ed4e5dce24d7652_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.57, 0.28, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63f4a6135a5ea4ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.75, 0.14], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8264cc8dcae09ba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.33, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_172c23b0449e0268_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fc5913112177801_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.26, 0.56, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1baef5fe234f253_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.11, 0.5, 0.29, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08e08ca1ec82e480_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.73, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87adb7078d351e49_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.54, 0.38, 0.07], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4703a48beeef821f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.47, 0.5, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5286466a7dbbbdf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.69, 0.04, 0.13, 0.14], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf590a676c77fa39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.63, 0.27, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d820d82dd055280_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.65, 0.13, 0.02, 0.2], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35297d17a8df51b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.32, 0.06, 0.56], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bbdf0505e4a71a1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cb8dccb3461dc0f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.47, 0.43, 0.09], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4abd0003cbf4a7bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01da22fd7035ddcc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.24, 0.14, 0.37, 0.07, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad062133d0c325ea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.2, 0.67, 0.06], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e955c40d91d5857c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.19, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0313e7dd945e2466_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90731cb2152555ac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.87, 0.09, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bcb1dd5f75574aa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.87, 0.11, 0.02], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_766aa1893edc931b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.07, 0.68, 0.24], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ba6b51a9d789ae3_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14e94b5daf0177db_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.22, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ed61424ba650d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b78a327f2b930281_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.17, 0.68, 0.1], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72d4871b874a4126_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf20fe92ac4a9ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.17, 0.83], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f3818bf941e5a79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.96, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_225e660bc56a39f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97cc7ba3218e88ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.16, 0.62, 0.21, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d138657f3f8cbe7f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.05, 0.68, 0.07], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c025d60ed07af741_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.15, 0.44, 0.29, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2e4be0ebf455f30_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.6, 0.06, 0.32], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23abde15345cdf9a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.13, 0.38, 0.42], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b76e24cb7638271_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.79, 0.18], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6e43fe666e20628_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29271f3582cbf5b3_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab3cb8e117d274a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5485a116a01bed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.07, 0.21, 0.46, 0.18, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc840a5700c38873_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.3, 0.6, 0.03], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d53c5f1c36e0c9e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.35, 0.51, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d967fc3b104c44a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.19, 0.07], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_663f119ca52fe746_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.6, 0.17, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cab5dcc30e13c52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.21, 0.69, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_206192a70f30ffc2_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13ad471089a036bd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.4, 0.12, 0.42], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cdb58f1055fa3c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.11, 0.21, 0.41, 0.16, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c17f0d955e8c034e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.14, 0.43, 0.32, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc99d808a355c871_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_990cef6bc329c560_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.56, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9efba9c1dee4c6db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0202, 0.1414, 0.3939, 0.404, 0.0303], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1261402c537bd438_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757d084c78c7a495_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.26, 0.56, 0.07], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3feee735ee025de_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_332242cd51ed0675_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.37, 0.47, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b01d579bfb13a303_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.27, 0.58, 0.05], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e3c541ad8f4c6bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2748b748dba9409_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fddcd871c406f308_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed1ba1bf71b18edb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.4, 0.13, 0.31, 0.09, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e303bb01e3307f3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.49, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e721bdaacd1596_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d967fc3b104c44a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.58, 0.39, 0.0], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_994f64a382cf1268_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b915dfdbdf3f7e90_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.14, 0.52, 0.29, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09ebf1016d26bc60_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.07, 0.77, 0.14], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6fa2d67bfee2335_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.22, 0.18, 0.39, 0.07, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba8108e3e0d7b1c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.2, 0.63, 0.09], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb0a363886e088a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62a423b47e1cb7c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.42, 0.55, 0.01], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f10fa7f1ab733145_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.24, 0.39, 0.25, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da58e7b489dd4554_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.25, 0.22, 0.13, 0.29, 0.08, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e33940d3dbb79e0b_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93181c923f545bf0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.48, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8370eb43558c3ccc_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4be6360aa7923d8b_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92020e4efe4f1118_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6246b31361574e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.36, 0.44, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77417bed6cf76430_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.3, 0.5, 0.19], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f5d81b477029272_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.05, 0.85, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65da0eb8fbba155d_n", "kind": "noul", "options": ["false", "true"], "target": [0.18, 0.82], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2102be411c2f55b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.37, 0.44, 0.15, 0.01], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ace53cfc46edd7a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.07, 0.14, 0.54, 0.21, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbf218dbc273ad2a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0808, 0.5657, 0.1919, 0.1616], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f58b7d223d8937a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.1, 0.5, 0.28, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aeaa72abfe290c5e_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a81efe2d6089cd13_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.66, 0.25, 0.08], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe8e23a9fc1bbeca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.76, 0.2, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63722a0fb1e1484a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.59, 0.17, 0.21], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883b3ca38a9fc3e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_552c7d8c66cb94fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.16, 0.2, 0.43, 0.15, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46234e5c609bfbfa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.47, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64d7d5879157dd9d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.31, 0.67, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b4539bb10a44f15_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.06, 0.91, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e955c40d91d5857c_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14f794220205ae2d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.11, 0.64, 0.22], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08bd7cfe88131a12_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.21, 0.26, 0.38, 0.08, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a57e2d0aff81f2f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.2, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa6d88b3e21fe30c_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d52ca6e5e78a8ded_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a381ea4de8b66554_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.09, 0.12, 0.76], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a95602dcc4285f68_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.09, 0.36, 0.5, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12b22ed7fc5446b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43696ef77c88edbd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.61, 0.05, 0.19], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75304b55a717170a_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_222c3451b5275f75_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.54, 0.4, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_448375f384e5f727_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.17, 0.41, 0.31], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ee10e94443f32b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5a2161acb694e75_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0202, 0.5556, 0.2222, 0.202], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fa0c5a2f0520acc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.22, 0.08], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fd57dd53243025a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.17, 0.61, 0.07], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41338733d8962b6e_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3dcf3fb0292189d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.16, 0.68, 0.08], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_355dd13532070f83_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.37, 0.25, 0.36], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8351234b700b0ceb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.43, 0.17], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a2b56812de9a4a6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.28, 0.63, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd357df2882a04a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56f5463daad1bcab_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bce82894511a1662_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.03, 0.31, 0.54, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e02d61ae92418c1c_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80d0920bff95cbba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.39, 0.47, 0.02], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6c664550416bde1_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f458844b670bcc5f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.33, 0.54, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_301b4b8c515c1ecd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.15, 0.21, 0.39, 0.14, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c603c56425afce81_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.22, 0.77, 0.0], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db06280c24cbb4bc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.05, 0.46, 0.47], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c995be57ffee36e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.16, 0.63, 0.14], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fed5de52c36aa7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.11, 0.16, 0.41, 0.16, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ee7760733d7a5e6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.25, 0.03], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be685001890a23e8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e00a6948ac35f8c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.53, 0.28, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_996f730d39e88438_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.01, 0.06, 0.49, 0.37], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c73ece668808d341_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.31, 0.6, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1859c2a7803f1860_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.38, 0.06, 0.46], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c7efdb32bc229a_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8662b4fa585b027c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.7, 0.22, 0.07], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cda5e0ac04c178a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc304ff26cacd3b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.09, 0.11, 0.48, 0.24, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88bcb9ec5fad216a_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_591c92803d1f2387_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.12, 0.51, 0.33, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8e2c13db052f0a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.13, 0.19, 0.42, 0.13, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87adb7078d351e49_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e35e68cc62497da_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47524887f080b56c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.22, 0.36, 0.36, 0.03], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_261e972ebb0362ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15934917b0f598e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.28, 0.62, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59f5b0ccec05f32a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.47, 0.18, 0.21, 0.14], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0446bac7b6cdd768_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.32, 0.62, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ab8a43409dd8c89_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.01, 0.18, 0.7, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b06a2c0e4f492a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.36, 0.53, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2309c64b94e9998_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.8, 0.08, 0.1], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73d24f5f3af5bd78_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.06, 0.84, 0.02], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99bab976b8a2fd83_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.41, 0.06], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_472a13f83a9de922_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed3ce46a025d62e5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.4, 0.54, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_961f2b8ac03fcf2a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.91, 0.08, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95bbd461edd2200b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.17, 0.38, 0.3, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80d0920bff95cbba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.3, 0.0], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d087f591ab7e31df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.1, 0.44, 0.37, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e29f1286a552c7b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.16, 0.51, 0.22, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ed5d894a365ff88_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.1, 0.48, 0.36, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d3bd4bcb1ecd027_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.25, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd3fd98514e3af96_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.09, 0.18, 0.42, 0.25, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c0c0713bc560bea_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df7011e46a340e3c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.04, 0.1, 0.31, 0.39, 0.07], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a35edc9ef07baa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.76, 0.23, 0.0], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5ea568950478a00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.22, 0.61, 0.09], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40da60d9b791a5d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c449d87bab1a6f8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.52, 0.09], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d57848d82b7737b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.18, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31502f6df0447055_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.36, 0.02], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa6d88b3e21fe30c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.74, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0ba375513d5c1e1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.69, 0.19, 0.1], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1872980fefc17814_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.17, 0.47, 0.24, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dff841a5fca40b79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.53, 0.37, 0.09], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea4ea17d3b05666d_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb46f9e08879ff17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.26, 0.62, 0.09, 0.0], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6647a8a6d9b5af4c_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67251e7c40941068_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.6, 0.16], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f1aabf902a63b59_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afb0412a7d68524c_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_802cd2a26e91ba13_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9986f2f3cfc5ca97_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06558d9fe1864672_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ae600ace3b488cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d81ca87440a3154_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.21, 0.47, 0.28, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e5085bd7582db3_n", "kind": "noul", "options": ["false", "true"], "target": [0.2, 0.8], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55aede0116108171_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.17, 0.7, 0.08], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e1a80b1ff390637_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.27, 0.46, 0.21, 0.01], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc384c67f2e9dcf7_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9108570beedd5a6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.01, 0.11, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bbffd9e161e322c_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b7d45acfa2750a8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.59, 0.2, 0.18], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab798e52d496dd5e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.24, 0.46, 0.16, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88fee1bbfc86fbbe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.12, 0.43, 0.36, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1baef5fe234f253_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.51, 0.33, 0.12], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a5b45d77bb52365_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8304af1aab219eda_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.67, 0.31, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0d0c4160fa6ef4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab5ab0b6e54a0597_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.32, 0.56, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3af0c9da946f9598_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.15, 0.55, 0.18, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63d5277473509627_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.6, 0.2, 0.18], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39080a730e4eb7b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.23, 0.52, 0.1, 0.15], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0654d5807df781d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.26, 0.63, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3368ead56aed3e7a_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35297d17a8df51b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6dbcdd59e70bef3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.7, 0.27, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c983db6e51a0f2d_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f615207b8ac41f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffac822c86e58786_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff4de9a138cd1a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.12, 0.73, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d55a0e9ea5d7c4a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.18, 0.24], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a41106a6764702_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.67, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4e99dde9304253b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.26, 0.65, 0.09], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08aa7edfc96aa080_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.29, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ed6d573c15094bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.68, 0.21, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eab9bd675bbbdc79_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9cf9e4b41b6ce66_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.23, 0.11, 0.38, 0.17, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f578d463866ceab9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.27, 0.65, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7951dd339505fcd8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.38, 0.47, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_924e149ba2fd20d3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.55, 0.21, 0.16], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01654f20b95b8a25_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bb40e0df477319e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.02, 0.07, 0.45, 0.33, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7e3b50d24648695_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.41, 0.07], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f12e76a8a4df476_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.6, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0007143882c4341_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.41, 0.56], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ea3b35b2df49a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d36cc6e958c1b3f3_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_975f0198d7d360bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d740bf72f15a8835_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b5eff964580e606_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.29, 0.07, 0.38, 0.1, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21de5fbb244ee77_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.3, 0.58, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6abeab9ee56a68b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.83, 0.05, 0.1], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35dedc4e441cbf9c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.06, 0.66, 0.25], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e9aaef111397c54_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38a9659a5f0eb094_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6869, 0.2727, 0.0404], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2ea9d3aa0abde85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.12, 0.47, 0.36, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_547c2427c3aee564_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.45, 0.1], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d245ec7e54735a34_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf3f02b36924b240_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21543e68a54acece_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5891a2ced68dec9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.55, 0.25, 0.05, 0.15], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5efcf89cd0f623d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.33, 0.49, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_072d65f96448fcce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.08, 0.55, 0.35, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcdbd4d1f8d5e174_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.6, 0.26, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2057bedcf356e9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5778e094db6dfb27_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91feaf377d0a755a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.63, 0.1, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66eeed9d5b1c6fec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.5, 0.06], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05971426306b782d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.64, 0.28, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40da60d9b791a5d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.43, 0.07, 0.28, 0.07, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5045901decfc8a78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.19, 0.59, 0.19, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_222c3451b5275f75_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.21, 0.67, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13b58a8f3368a0b8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.11, 0.48, 0.29, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_222dec12a2dc24bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66d77e5905e590bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.5, 0.41, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fbc3c94fae39a8c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.08, 0.38, 0.4, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277bf9874476a3aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ae43a6a43f7f434_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.18, 0.5, 0.22, 0.01], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dcaf766b33254f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba1e862ebb5fa056_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a2e6ad0e07216e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.33, 0.02, 0.39, 0.26], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c21709c7047ec6f_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b08100c3159e914_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.13, 0.22, 0.47, 0.1, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6071d1f8adfbb87_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.08, 0.91], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33f7e0fb5b177a50_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_478b6646136138e3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.04, 0.77, 0.18], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd9e3f7b60f05a18_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e56d42be7b345ed9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.11, 0.44, 0.36, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bcb1dd5f75574aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.3, 0.55, 0.13, 0.0], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66b4f5086c99afff_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a84e85eb988ff68_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.88, 0.02, 0.06, 0.04], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_777e864c48b763f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f2da9288a69331e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.18, 0.48, 0.27, 0.02], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0dc35d3ebe9f750_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.28, 0.54, 0.17], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9bdd34ecc6fc334_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.09, 0.61, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cb8dccb3461dc0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.33, 0.55, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e0349440416b9d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d998f63507f7b47_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.19, 0.5, 0.14, 0.01], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bbe20ef5ebe63435_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0303, 0.2323, 0.6768, 0.0303], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37b280193138c49c_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1217cc4e8c6381e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.27, 0.19, 0.32, 0.1, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37c7572491a56f7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.12, 0.77, 0.09], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2057bedcf356e9c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.79, 0.1, 0.08], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a8784d29d315a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_552c7d8c66cb94fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_155492440990c9f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2ea9d3aa0abde85_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.75, 0.22, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d6709f41072f3a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e8cb07fbef09878_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.24, 0.11, 0.58], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5804b1ea89d925da_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa072517cc2732c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.25, 0.48, 0.2, 0.02], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6c4d58182afc6e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.27, 0.2, 0.33, 0.09, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93814c9337a9b736_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.03, 0.88, 0.05], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8872302717cfdb4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.41, 0.01], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cebe7973f99e3acb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.35, 0.58, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a491e6482550bbef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.41, 0.48, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_385ac11f5cb83867_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdf4ae97bd7f98df_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.48, 0.28, 0.21, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45d65f9f52d26978_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b820ff4de9246dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_133d70a53cc0117a_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855b343674aa5c6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52560ece42236d79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.4, 0.34, 0.23], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_994f64a382cf1268_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.47, 0.51, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce43627122618cdc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.38, 0.18, 0.27, 0.04, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5209a026db1a622b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0404, 0.0505, 0.6768, 0.2323], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21000fa3ed8bd88e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.3, 0.62, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b17db984026b6c04_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.18, 0.25], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2eb3a3950b3e1b34_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.09, 0.58, 0.18], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5268a769c88ab982_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.28, 0.19, 0.36, 0.08, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51f2fa5f7db0b5ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67251e7c40941068_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.35, 0.53, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72c3f40d6f20384d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.32, 0.59, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_596e0baf008eb976_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b25e14c861f3e393_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.03, 0.24, 0.62, 0.05], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7308bf3f0058c13c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.23, 0.56, 0.15, 0.01], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4fcd1f6c69d31e1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.77, 0.18, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d472f467babd2c1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.72, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ed61424ba650d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.2, 0.12, 0.36, 0.19, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1e3611a90fa1ecd_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0de791aa3609e92e_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eaa209b839a4a32a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0101, 0.0303, 0.3131, 0.5657, 0.0404], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9aa654dbc67977a2_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8a881063e65d1f0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.68, 0.08], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ca067c8d7da4c0b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.1, 0.35], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fd1e0ac8834f163_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8304af1aab219eda_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.19, 0.75, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_217125549d0ef849_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9a28e56e8d45212_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_529aad22c700fd6d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.63, 0.1, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb51d311c2f1d68a_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d55c6dd59b69ddaa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.51, 0.34, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac58503902688d22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.32, 0.04, 0.49], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_731695004796ed3b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.57, 0.27, 0.1, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a9e977023f1a4e3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.72, 0.21, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_996f730d39e88438_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eed868b0f0a5eb9_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01cd0da562c8c559_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_919ed2d2f3055219_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_decd5f546f09c6de_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_858aa1a2838e1f38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.18, 0.45, 0.29, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aeba1cb3e4ea934_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.12, 0.48, 0.34, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c366bb8423e6d9c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.16, 0.53, 0.18, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bb40e0df477319e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.23, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c440e4bfdd26bf2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.71, 0.15], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_480f974ff7a0689a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.17, 0.39, 0.28, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6c5d1498f6c4a4c_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3f7d0e8e4ab1e20_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.22, 0.64, 0.04], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4da32cdbca57989_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.06, 0.46, 0.48], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e061b301f3f14733_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb6ff29f04e50920_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.34, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c0c0713bc560bea_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.33, 0.1, 0.44], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70f55cf635c50299_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.09, 0.28, 0.5, 0.09, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eb7d16210199b8c_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56d1c8335e745f07_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.77, 0.18, 0.04], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8392d13d9f2054c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.11, 0.48, 0.31, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_592823c8399070df_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.07, 0.79, 0.09], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc97c5ea9f2e2409_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.04, 0.4, 0.5, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f35ec43466addbfa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.91, 0.06, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff4de9a138cd1a4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.54, 0.11], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54121102ffa889d4_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_547c2427c3aee564_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.32, 0.56, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f6795aefb4a59e8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.2, 0.26, 0.52], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415f43981ccdcae3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.08, 0.23, 0.5, 0.11, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33d3dd2a6643c3f7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.56, 0.04], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d6aa770e70f57ea_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.49, 0.14, 0.36], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aec8bd310383019d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.39, 0.12], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_707532d73f7bbbe2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.22, 0.34, 0.21], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958385dfbf30315c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.11, 0.37, 0.4, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958bd4d1642f3f2c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.9, 0.09, 0.0], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a96fde324f2d8f87_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.6, 0.1, 0.21], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f36243997e53b9d4_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f021521c533974d_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f89c97182a70660_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db72b6b29fc32cae_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd693af83ab957e6_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d43d0b7bc23d5340_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.52, 0.42, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a09946e4e742354_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.3, 0.04], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fcc88289efa28377_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.08, 0.73, 0.15], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a542802f05a29c80_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.5, 0.35, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b33c9af93a60ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.19, 0.27, 0.44, 0.05, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_355dd13532070f83_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7222b0ea154d4d3c_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da1a7e66012e69a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.94, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccae4b6c9c3c4d3e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.3, 0.64, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09ebf1016d26bc60_n", "kind": "noul", "options": ["false", "true"], "target": [0.98, 0.02], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97e27660e51feeec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.28, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bcd316b7e63330a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.34, 0.44, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a4a0a4e6c7e64e2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.73, 0.17, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de0e925a5a4ea8fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.24, 0.63, 0.11, 0.0], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_162c85768348fb98_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4761fe14f277aab3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.09, 0.46, 0.34, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f470afbe78692227_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.0, 0.01, 0.68], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca98020acc81c7b_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c76677ade9a3f58f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.28, 0.59, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8f3681f6c1193b2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.94, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e24e2f5ea3157d6_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3073d1798202acd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.86, 0.12, 0.01], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_661bab7f5225c8fa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.39, 0.06, 0.53, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6cc706f07b50eb23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.77, 0.1], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4767294664e19455_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.15, 0.74, 0.1], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fc04f26d489b0c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1f96312d0f0eaf4_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e5b50b9358a2ac1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.34, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f458844b670bcc5f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.18, 0.2, 0.54], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c10747ffb2b406_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd791bb12d3ff35b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.31, 0.16, 0.33, 0.05, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdd1b49bc922fd76_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.38, 0.6, 0.01], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87e23308981e4f35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.1, 0.77, 0.11], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7bfe123480c81a5c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.64, 0.24, 0.01], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12f86d3bcd3b66ab_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1c3e2fb1d4a683a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.03, 0.93], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa316cc43372027e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.74, 0.21, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d43d0b7bc23d5340_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34b520f7f5c47e07_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_752aa953c24b7b17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.08, 0.77, 0.14], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88ff11b101851bd3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.29, 0.17, 0.34, 0.08, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83e6214fba4ab02c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.49, 0.3, 0.02], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4575dd882b0c462c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.16, 0.48, 0.2, 0.03], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f76fc6c53e5831a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.41, 0.54], "state": "An automation plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e43971168e23c79_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0386211b077ff799_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.41, 0.5, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_767c6dc95eeb6fcd_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e071f06d236c9a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.37, 0.11, 0.3, 0.08, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_644841e296a23901_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1ed3b6f9d499ab9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.16, 0.47, 0.25, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c025d60ed07af741_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.45, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fa899a358423b86_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13c8abd3e6749b00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.11, 0.43, 0.39, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08dced6b7d058edc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.0, 0.01, 0.63], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1274b5e329255a8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.54, 0.18, 0.03, 0.25], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6fd47ee23167d52_n", "kind": "noul", "options": ["false", "true"], "target": [0.19, 0.81], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cd29498fb8de605_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.39, 0.5, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9dedbf61c2cafe72_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.04, 0.9, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8baa60c91ac64306_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.43, 0.43, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6c9d1519ddf3c8d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.37, 0.12, 0.32], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0e31868721b782_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.53, 0.17, 0.26], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8a4188f07f4d3ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a064f56facd998b8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.95, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_175ae805e63738ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.14, 0.59, 0.19], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_111f2f1ce23ab7c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43c4c231c0b289fe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.22, 0.32, 0.37], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f0414f57d29acf1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.14, 0.45, 0.25, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebb7dbc52e7cb1c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffac822c86e58786_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.73, 0.08], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7cb3d9f6bb1930e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.1, 0.5, 0.32, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cab5dcc30e13c52_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed6f4e25a786ceba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.5, 0.05], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_122f793ad020a6e4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.85, 0.11, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5549ccef0870e06b_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_718a22a42ac21bcb_n", "kind": "noul", "options": ["false", "true"], "target": [0.24, 0.76], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1859c2a7803f1860_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.22, 0.22, 0.42, 0.06, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ce3b9d955a7440_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af4dd96e411d8c26_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7053a11502fa67ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1504b013343c317c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd11ff0c5bb4dc35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.2, 0.68, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2864e21eea12f295_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68312a8de10982c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.1, 0.39, 0.43, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a439926d35bddc7b_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e0349440416b9d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.17, 0.11, 0.29, 0.23, 0.14], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb4aeb00f54597b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.07, 0.78, 0.14], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_267be4a45109324c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.19, 0.54, 0.2, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf6ce920c3ad4c11_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.78, 0.01, 0.09, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34525ee4851c3efb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.51, 0.08, 0.38], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05219f41b5403ab2_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7e55db04f26ca82_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.37, 0.09, 0.33, 0.04, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_067018c50b7b46cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_708b0c3ae85e9cbc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.41, 0.37, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a82bf8711f7925b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.6, 0.19, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af9273905b1d6f59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.16, 0.68, 0.11], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a41106a6764702_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.73, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43d1a96eb9414e7a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.03, 0.11, 0.33, 0.39, 0.06], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d22f1ffc61c3652e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.09, 0.13, 0.46, 0.19, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_184a39cb430fff95_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.1, 0.39, 0.42, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06ebc2dbe9754df8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.16, 0.33], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98bd266dc9ba6a31_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.28, 0.54, 0.06], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_622a3b2ec7c5d457_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca9a4252b1c46c8_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7899499ff84a8e85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.13, 0.58, 0.19, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51bc82039a5bee02_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11a283b71f20ba17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0808, 0.1919, 0.5455, 0.1414, 0.0101], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73d24f5f3af5bd78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.28, 0.46, 0.24, 0.01], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f087bcd909fb94_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.03, 0.06, 0.88], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_097b94e33ccc4a83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.2, 0.58, 0.09], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56108d3783f302f2_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_047f4022d0ade92b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.09, 0.78, 0.13], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3cb5d98b1cd02f9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.23, 0.67, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cab5dcc30e13c52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.1, 0.11, 0.4, 0.28, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14c387dad35e6ea5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.55, 0.06, 0.25], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8806d4268e4227cb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.22, 0.08], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f64aad792c95d5d4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.44, 0.22], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44fc15661deebd59_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.94, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_261e972ebb0362ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.49, 0.37, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7efc135282c7cdd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f53bf6e414d7c816_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.81, 0.14, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab6934e360225844_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.02, 0.11, 0.62, 0.18], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_519d267c915a30ba_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.03, 0.92], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d59ebad8df3fb4f_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edb14bb1e22c861a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.08, 0.35, 0.45, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb87d018c2278c4b_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_421aa151f47e2594_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.68, 0.29, 0.0], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38693aab8bf70b9a_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9b1e527b5b4d995_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.25, 0.2, 0.32], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a529a702863b05a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.22, 0.19, 0.38, 0.12, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a533e70855df6cac_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1090496fd73af91b_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca29bbab70b45150_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d98be830004d47cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be7fa7c56f2be55a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.22, 0.08], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c96233c729a9e5e_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfed58128702df71_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab040674edacd800_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.88, 0.06, 0.06], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a1d403c0f5d2ae7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.09, 0.41, 0.45, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20a3331b7a65035c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.07, 0.17, 0.48, 0.17, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98c01fa5c0ae5ba5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.35, 0.11, 0.24], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98db039c333707d0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.97, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933689cf11261639_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.29, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b16b6853a83605ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3cd9db5dd8602125_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.34, 0.6, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c083cc41faadb72f_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_405ee0a1961c919c_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e79d165c8e87629_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_353dd2c1fdbcb8fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.11, 0.72, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92ecefa8b68abc93_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.11, 0.46, 0.29, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_feec5a816ad8c4c3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.38, 0.18, 0.39], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31d31d3b5532172f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.12, 0.29, 0.39, 0.08], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ab4cd676c0d9d4a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.66, 0.31, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a239a2045ad73b6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.63, 0.05, 0.01, 0.31], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f1ef042f62760d6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.2, 0.07, 0.71], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44e3d31704978752_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.18, 0.69, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_110b19e40ad41be9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.26, 0.61, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d886a5259de93355_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a57547febd908917_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.16, 0.5, 0.2, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c169b34fefabe23f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.17, 0.13, 0.26, 0.27, 0.13], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d79e03b47aa640_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.67, 0.26, 0.05], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d39c1722346d1e3d_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2e22c8aa30ac982_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.09, 0.68, 0.18], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7d9ca5f042e8a9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.06, 0.59, 0.33], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5235c825e7da5452_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.65, 0.12, 0.07, 0.16], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e387340aafa3246d_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9944f84e1c8569c9_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf3f02b36924b240_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.39, 0.55, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a39aa2ed422e7089_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5f56fab736f4fd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_880905eebc7f9186_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.27, 0.43, 0.24], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d97b70365c031da_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.73, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_736d4923cc9210b5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.27, 0.52, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fe247362fe3b19c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.25, 0.62, 0.06, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5ae31fa0847e941_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.23, 0.66, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ef28fe276a4fdb0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.57, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8fd7ad4d05b8572c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.13, 0.72, 0.1], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c00a252f79f34100_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.46, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_172739dc8dc5d65a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.06, 0.45, 0.31], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47ad3b788abdbe68_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0202, 0.1111, 0.5657, 0.2525, 0.0101], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea00f02b194aa3ba_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.24, 0.69, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2169f6bae8231df9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad22aad8c95dd3d4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.28, 0.63, 0.04], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b16b6853a83605ba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.49, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6fcab09f8d51e2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75e882b5549a9e65_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.36, 0.52, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61c0d40d4c581382_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.63, 0.17, 0.02, 0.18], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbfb49fc3e225d22_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4c54eee3162a935_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.57, 0.43, 0.0], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d619521db9741089_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.04, 0.26, 0.55], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ed5d894a365ff88_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39080a730e4eb7b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.1, 0.19, 0.43, 0.18, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f768660d93aaf24_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f0589a1b10f30e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.4, 0.23, 0.34], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cac519fc57cb0e0d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.21, 0.55, 0.16, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff4e779b8d4f50c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b41017e859db61aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b637ac7d0986ce6_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_505ce5360250384e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.15, 0.45, 0.27, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_982a3f76b092803a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.37, 0.54, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a0f395149b8d83c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.53, 0.37, 0.09], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53aa6f13658fd53c_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2d59e9a56cbd568_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.19, 0.38, 0.27, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca3d7179e438f446_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.46, 0.41, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54fb03c319d86867_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.45, 0.52, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a6321fed1eb5486_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.41, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee340f3175e3ca3a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.05, 0.1, 0.37, 0.35, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f4073e8d1b0d24b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.15, 0.54, 0.17, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2d59e9a56cbd568_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.42, 0.29, 0.24, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2a205b636942a70_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.46, 0.4, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_674fa20a6ef65124_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.14, 0.65, 0.17], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9333317779ec4a8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.58, 0.02], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_228949039af8c595_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.42, 0.2, 0.36], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_555c2537120e90ff_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.59, 0.1, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58671bcb7cbc84fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33046db768a0b74e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.05, 0.92], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1306acffc48aefb0_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fd882fdc8921310_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.18, 0.47, 0.32, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9488f56f4cbc6921_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_707532d73f7bbbe2_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b04d58054fc1a12_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.06, 0.65, 0.22], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b432480ee5d1d219_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.1, 0.18, 0.42, 0.16, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40177b253a9b801a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.07, 0.49, 0.35, 0.02], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b04691f0b252d89_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.28, 0.23, 0.3, 0.06, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d88b973563b960f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.18, 0.49, 0.14, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68312a8de10982c7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.49, 0.31, 0.16], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12b6d3a797355ef2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.3, 0.54, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2a3caa99a4fe6ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.07, 0.92, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffe1ec62c293df22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.21, 0.17, 0.38, 0.1, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57951842bd33e103_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_061c05b4e5348ae0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.14, 0.33, 0.38, 0.07], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6193b0bc82dcfba8_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eda7f78582fe0d1a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.52, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bc8b24a2f29e037_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fa2caa67bf70886_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_000317cb5bc25cb6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.67, 0.08, 0.21, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4941af120a72dd48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.09, 0.29, 0.51, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56f5463daad1bcab_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.28, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d967fc3b104c44a_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5776e09ffb9c0feb_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_789df7b599d58676_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e228a5203bbec4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.31, 0.55, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10ad3427799710e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8025f137d326edb7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.38, 0.08, 0.51], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b949182670528683_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.91, 0.04, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50329125c3824eac_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e04db42af010212_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.25, 0.16, 0.39, 0.09, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ef921b6e17feaeb_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A proposed workflow assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba2b7f5e0983a194_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51a85ba0828b8441_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.29, 0.01, 0.03, 0.67], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26a1f0dfd25aa931_n", "kind": "noul", "options": ["false", "true"], "target": [0.23, 0.77], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03093030590dd0ae_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.02, 0.66, 0.27], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc5e620a1d4b512a_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ea69d1f9f3ec006_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.35, 0.38, 0.26], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86cbef0dcae9ef66_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.65, 0.17, 0.08], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc3b2faa809b7b4f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.2, 0.8, 0.0], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d62d1d6eb099b4b3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.2, 0.75, 0.04], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba20009e5f08de00_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.75, 0.21, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1cb6cc228707ade_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.71, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d4bd6730b08629e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.07, 0.1, 0.11], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7369bfb8e3e4fdfd_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45a25f691bac9246_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.61, 0.29, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75e882b5549a9e65_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d2b96c08abc7ec2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.06, 0.49, 0.43, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1f3182d1df599c3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.29, 0.29, 0.14, 0.28], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b946f6faa72af7f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.35, 0.06], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35628e238adacf58_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_568edfd33eb3427e_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31e1bfdefc81e199_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.16, 0.8], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22892e44c523bf86_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fa80c99d27ff915_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.14, 0.42, 0.38, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d86880153e748db2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.67, 0.1, 0.04, 0.19], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10ad33721f8bf53c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.25, 0.47, 0.15, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f96d485ce9b29d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.14, 0.5, 0.29, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dba04db8b39179a_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d4ba213ce853c18_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.18, 0.14, 0.41, 0.12, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ab7589464b00097_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.63, 0.36, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f47db1c0d1769609_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6561aa7f547dce25_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2828, 0.2626, 0.2828, 0.1717], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_818e6a5912f415bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff30d71d81f121bb_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_719882f82c0d1749_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.79, 0.05, 0.07, 0.09], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f088c691a359dbd_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a180bbc6105b776a_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61ae77c3e1230d9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31dfc80f7da53384_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c11fdefaeee82fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.65, 0.15], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59f10befe0e179d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.53, 0.39, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a9bf3f506759baf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.24, 0.46, 0.25, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0bf9390fad569749_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.79, 0.11, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b879ae8191c0d6d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.34, 0.53, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757ec56d5379abce_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c46a11442eeaf769_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.06, 0.7, 0.23], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0ab3d681769ef13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.24, 0.56, 0.15, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8671fe913affe906_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.2, 0.65, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e326816e61cdd0a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.08, 0.23, 0.49, 0.16, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d07625334564b51e_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44cf965567366988_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.5, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39872950d2be951a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.1, 0.17, 0.35, 0.2, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1acb8a698b381b62_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d117fb9c4159689_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.89, 0.09, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88fee1bbfc86fbbe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.7, 0.05, 0.22, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff2c14058f2a4699_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.76, 0.19, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_810ed800d8ef1135_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.3, 0.55, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_322cc44ab44ce591_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.02, 0.22, 0.71, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb29893d55dccaeb_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0be14e0f640b501a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.56, 0.24, 0.19], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5ee7b10138ad9b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.22, 0.65, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a0e33f68fb35503_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.7, 0.1, 0.18], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1c4766d6a22c512_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.17, 0.58, 0.23, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad4ddb9c5f163744_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7feeded03d46b560_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.49, 0.41, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb51d311c2f1d68a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.0, 0.98], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b17db984026b6c04_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2223a944aeb00b0_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_236be50aa8a0bd05_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.23, 0.72, 0.04], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9c40fba04034cf1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.84, 0.13, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f3818bf941e5a79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.07, 0.5, 0.37], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de405fdf4136611f_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b71afc446c0a597c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.04, 0.24, 0.36], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2949f6caf642b23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.48, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ec9e05f6d541467_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.14, 0.41, 0.23, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27845f38518fe091_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.5, 0.04], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_135f3b0f5b9a210b_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d35e59f8edd8f5db_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.73, 0.01, 0.09, 0.17], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaa35ebf336be0b6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.23, 0.35, 0.3], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c440e4bfdd26bf2c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86a2054841463def_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.91, 0.06, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdb7f65c07475cec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.11, 0.15, 0.46, 0.16, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9e961fb3ed33e3e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.5, 0.36, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0707c19aec893dff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.13, 0.21, 0.4, 0.16, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f97723a78b20cf6e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.49, 0.37, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3597e69e3d0fcb37_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d79e2a803ffc60a3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.23, 0.52, 0.17, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdaed520e703c97b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.21, 0.56, 0.2, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b9bc07380cc09b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.58, 0.17, 0.22], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ec4dc64d0f40041_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.14, 0.81], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1056fb002cbe3e62_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933e1d34a769fa51_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.71, 0.01, 0.11, 0.17], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd0f37338c2f6654_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.23, 0.2, 0.33, 0.09, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87bed3086254f7f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.05, 0.15, 0.57, 0.14], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d018716c83bc711e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.09, 0.29, 0.52, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e41a425416829a7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.05, 0.95, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0e022fb3d28d283_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.01, 0.11, 0.64], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4743134b82746ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.19, 0.7, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9efba9c1dee4c6db_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3f7d0e8e4ab1e20_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8fdbe9c3a61c09a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abea6d3f20a042c3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.05, 0.91, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9873a983b7c85a74_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.12, 0.39, 0.4, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3978ec1143b544e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.65, 0.13, 0.01, 0.21], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63d2c885cbee05bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.71, 0.07, 0.17, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e38609392eeea46_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.3, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a9b344c3038a211_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44b186cbc8a235b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.15, 0.69, 0.09], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9016528f9181be68_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.68, 0.04], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a154c1c7866604de_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.26, 0.49], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1dbc7b4473888c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.44, 0.44, 0.01], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c88659e68e011d37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.18, 0.7, 0.07], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6379967d2da0806_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.19, 0.19, 0.4, 0.08, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_760b68fbf64a62b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.27, 0.6, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_825f1de029b6647e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.25, 0.61, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80aa4a56a58b0553_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.33, 0.55, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e55208dbc6434914_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.43, 0.38, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0b1ee898b4a1379_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63722a0fb1e1484a_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80b2a5bb76171bd3_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cae845aabed7648_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.23, 0.43, 0.26, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3749e6f4bb7c9778_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.85, 0.11, 0.04], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db6e1421cfbc47ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.15, 0.65, 0.14], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1afbe6ef6d34f17_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.33, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9060c353303d6444_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.27, 0.59, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d7e4f38c531b1c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.14, 0.37, 0.3, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d652fe35fc8fb541_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.06, 0.14, 0.42, 0.25, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a219998908c51706_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.19, 0.51, 0.21, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d81ca87440a3154_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2748b748dba9409_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.22, 0.33, 0.42], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13799d7fa88ab187_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.17, 0.53, 0.22, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c20e8f9be4102b94_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.63, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89783f6a5bb585c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.19, 0.06, 0.64], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52560ece42236d79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.52, 0.37, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_846cf85124d7c59b_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_128a5de4f81774f4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.44, 0.45, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5be4beaaf631929_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cff6dd29686864bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.04, 0.06, 0.26, 0.49, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c4fecb78131315d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.88, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50b71c9cabc2f712_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f4187afc6f3ed73_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a554dfbe81015a5c_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67bbceeaa11de3a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57791e80ef540d39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.27, 0.54, 0.14, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c3eaf46f60b44ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.14, 0.38, 0.44, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13ad471089a036bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.09, 0.44, 0.31, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be7fa7c56f2be55a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.05, 0.42, 0.52, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fe84315f1ee110a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.3, 0.36, 0.2], "state": "An automation plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_846cf85124d7c59b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.26, 0.71, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c16bc831c9fd5cd3_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f079978f2f8eadd7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.62, 0.3, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6166e74e86e271d_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b6d4346390bbc10_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90915e2a9c79ca9e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.16, 0.51, 0.15, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b9515c4826bf534_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.38, 0.59, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5268a769c88ab982_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ac6f9c5a9b9a09_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91b615df4bd0f02e_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_357f8f38c41b86f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.07, 0.37, 0.46, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5860f25b7abcede6_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba32c0160d76d762_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.16, 0.43, 0.35, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_559773473da5fd9c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.55, 0.08, 0.3, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4048346a3001a13_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "A proposed workflow assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2a82438b5e028cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d224c3c24878ac7d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.58, 0.07], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1893c5ae8a303aa8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.33, 0.0], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04ec83834058a5c1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.33, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc1c960b7d07b520_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.41, 0.16, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9bd2d7ead6a2904_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.5, 0.06], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee340f3175e3ca3a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.3, 0.19, 0.49], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ae658cb01f7a0de_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.31, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afb0412a7d68524c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.5, 0.38, 0.11], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14a01012c8c330e3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.59, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40da60d9b791a5d9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.32, 0.41, 0.1, 0.17], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f97723a78b20cf6e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.43, 0.41, 0.15], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_975191f5cc8e59fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b3eabf7b7235eee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.29, 0.08], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_234a075eae0575ac_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.78, 0.2, 0.02], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48eec294d5fed903_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5b71794debfe7dd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.06, 0.64, 0.26], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_569000e1f3b19cf3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.61, 0.15, 0.03, 0.21], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c20b6926910828f_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14f794220205ae2d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.15, 0.52, 0.26, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9387238920d9c009_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.05, 0.23, 0.59], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_707f6746805be952_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.17, 0.19, 0.42, 0.11, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ca8063074990bae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.34, 0.12, 0.33, 0.07, 0.03], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2267478b77d38bff_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b610cb03c2449cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.38, 0.47, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_634cf83132e0a41b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.11, 0.51, 0.28, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d52ca6e5e78a8ded_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.21, 0.69, 0.06, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e41320686c776c89_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.35, 0.64, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66d2495125e6d86c_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7cd5b69df616bef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.39, 0.14, 0.3, 0.06, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bf784fbd3d66fa0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.1, 0.75, 0.13], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cbc861a2205b8081_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ada55d12ee565707_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff682b592e24c80e_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7504ae3bf5d8443_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.18, 0.71, 0.1], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4bb5719058fb682_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.77, 0.21, 0.01], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4428f4cb876dda16_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.23, 0.69, 0.05], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5dc9c1a5d0de92e_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9cb6eda676554d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e4fc6ddb6691329_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.2, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fff8d5e8907ea1cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.41, 0.49, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69d438a3a623f30d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.03, 0.09, 0.32, 0.41, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_864505e3705ebc96_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd18c81a633c8160_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.01, 0.96], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca2037718b2c525f_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac1b5f3ac7841ec1_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_effb9fe937e0e689_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.27, 0.07, 0.59, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d1b3d91e66e236d_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9c40fba04034cf1_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7589675176022f36_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.2, 0.68, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7bd529620ea014f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.48, 0.15], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1b18fefa2c73a1d_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bbffd9e161e322c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.23, 0.61, 0.12, 0.01], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a09946e4e742354_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48a77b7e4ae53d10_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5f56fab736f4fd6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.05, 0.8, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e206d508c217e03_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6ba799d568c4552_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c7ec41836541945_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.17, 0.42, 0.26, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ec73a7d425b6710_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85d3816d059cec2f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.12, 0.75, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dba209889718b3ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.09, 0.17, 0.46, 0.15, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f1ef042f62760d6_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_528bb034edcc314e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.14, 0.19, 0.41, 0.12, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aed1c184fa07bb3e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0202, 0.2121, 0.5556, 0.202, 0.0101], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74487a1b946d1919_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.18, 0.46, 0.29, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e24576aa6157657c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.36, 0.53, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d3d3cb81ff91d34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.12, 0.31, 0.49, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8292c1f469b0ad1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.38, 0.51, 0.02], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e1f7a2109ed89c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af0007c0abde98fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26dbbeb886f3f1e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.31, 0.51, 0.11], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_504e7f1a770f7fcc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.22, 0.74, 0.03], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11fd7eacb93f69c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.24, 0.7, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_544f966967942aee_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4cfbe1e7693abbd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.3, 0.67, 0.02], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a40ec5dd180fd28_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.68, 0.09, 0.17], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea2d938aeec9c4b1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.73, 0.08], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db8c9047f82fa366_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.22, 0.64, 0.08], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6459d294028ee5c_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0654d5807df781d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b89f315163b792d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4bfbf7f9f2686e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.17, 0.62, 0.12], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d39c1722346d1e3d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0404, 0.1515, 0.5556, 0.202, 0.0101], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdae166532f73e68_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d946e6895c48574e_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f84f1ad928eae97e_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d5ff0e9c95c65d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.14, 0.49, 0.32, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35eb8a38e01d39fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.18, 0.46, 0.24, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_826e26ecaa12e4ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.09, 0.67, 0.18], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce8d2b8077eb5e54_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.25, 0.12, 0.36, 0.15, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9ebfaed888061d0_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23f0d1c8afa48565_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e943f99a67f71940_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.13, 0.83], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b8da37c3cb5eb6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.41, 0.22, 0.08, 0.29], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415da395b7b74348_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.27, 0.6, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5d8eb7c65766023_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.15, 0.42, 0.31, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b04d58054fc1a12_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59f10befe0e179d5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.38, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b2124149073c7f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.29, 0.58, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a154c1c7866604de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.36, 0.53, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbd7380eb16fc26f_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22a69990af1c557f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.12, 0.75, 0.11], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64eb30296601efdd_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_462058791d0c75c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed3ce46a025d62e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46e25b4310197415_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.35, 0.54, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d9af36e945b5b68_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.54, 0.44, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a35b447793ac5a8c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.17, 0.68, 0.07], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9051a131e1e9333b_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_406dc854d781f9d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.14, 0.65, 0.15], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29d3c5e68fee195d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.77, 0.18, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277212b08de14879_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.42, 0.19], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7a2244760cf3c3f_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96f77196de8f2658_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.89, 0.08, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6335985efefdf4ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.27, 0.64, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0efb6ce86845dab_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2242e6693ae6033e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.26, 0.58, 0.15, 0.0], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17abb1ede5c0ddaa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.37, 0.16, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87bc98fedd0d714a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bd6c5317d2d45d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.18, 0.76, 0.04], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_003d321459de0e7f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.16, 0.81, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3da706bde7c5d4f1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.32, 0.59, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff4a76b023bcf36_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.42, 0.05], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02cc3d64951323a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_00d5e20ba0d2e893_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.43, 0.1], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2aa81e39eb618cc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.88, 0.11], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5f478b8cf56baf6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.11, 0.24, 0.47, 0.11, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8eaf138322cf77a1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.04, 0.43, 0.44, 0.02], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35eb8a38e01d39fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd1c0a90154413de_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.09, 0.41, 0.34], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_333992169e79c9df_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a94bedb65a50312a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.76, 0.2, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aba5f3908086fc5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.68, 0.16, 0.14], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e8f99d6b0cbcc90_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc116055dbad736f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.58, 0.1, 0.04, 0.28], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_489b4b3596340e77_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.84, 0.1, 0.05], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec4f9151be116809_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.02, 0.06, 0.69], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4082dfb31a3f74e_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3d915ee67f10c67_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.03, 0.06, 0.35, 0.44, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4286f67cf12195e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.36, 0.57, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b957f2f047e50c4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f768660d93aaf24_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.22, 0.26, 0.06, 0.46], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e35e68cc62497da_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.23, 0.67, 0.05], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29e1dbe1671bdd21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.14, 0.67, 0.11], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41a5483da99db0fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.15, 0.47, 0.18, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86990f33ce63e7a0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.14, 0.77, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_511c24ed4e9a3d22_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efc7522ddc465e6d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.34, 0.53, 0.02], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b5872bc773183a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.02, 0.03, 0.13, 0.52, 0.17], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e364f2d6c3b685a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.08, 0.02, 0.84], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_116b28e301cdd6d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.36, 0.49, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c98aafcc5e4fb11a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0, 0.0505, 0.5657, 0.3737], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e38911b976da5cf6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.23, 0.74, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aa3991616910d45_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.09, 0.2, 0.49, 0.17, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5edb579c7d7a6b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_178fcabbbb427a4f_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea5fbf11984ced73_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.4, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87e70297987fc237_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f0deb9bd988c07a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.61, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08e08ca1ec82e480_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.02, 0.25, 0.62, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c49ac7c6a3c733d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.21, 0.62, 0.08], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc116055dbad736f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.5, 0.11, 0.2, 0.05, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ade2e6560f3af5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.32, 0.19, 0.31, 0.09, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1109e15f85251d89_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_421aa151f47e2594_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.35, 0.01, 0.15, 0.49], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cec3df879b9b1b73_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.5, 0.32, 0.1, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56719963804128a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88ff11b101851bd3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.69, 0.28, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db8c9047f82fa366_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.94, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c2fbe5cd2f83545_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0007143882c4341_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1b7daf75bad78c0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.25, 0.51, 0.16], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf6ce920c3ad4c11_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.44, 0.48, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ee15a08a5c44163_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.44, 0.51, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5dadb961d148c1b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.48, 0.01, 0.01, 0.5], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c3eaf46f60b44ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e750dc810a4bf277_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f728f2b26b7fc214_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.03, 0.08, 0.26, 0.5, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26a518b1f108c42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.21, 0.48, 0.22, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9fdafc77124012f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.34, 0.17, 0.28, 0.06, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e4210e866273a7d_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afcc6160291fb4b8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.24, 0.65, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_928e62f67aa8982b_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd8bd91b2322284d_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2999fedb4e0322b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.6, 0.09, 0.24, 0.07], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d76834f14ae6d84_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.3, 0.17, 0.51], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f94e1158f0d9a6b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.26, 0.31, 0.4], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c539d91eec364a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.15, 0.17, 0.37, 0.16, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08aa7edfc96aa080_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.15, 0.51, 0.24, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68f8c01fd4740486_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.18, 0.17, 0.39, 0.13, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_184a39cb430fff95_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.33, 0.0], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1b9064409908d7d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.25, 0.69, 0.05], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_deb5e26222a441e6_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_128a5de4f81774f4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.07, 0.58, 0.21], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f9cef3e41a6bb6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41818955047ed9ce_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.37, 0.45, 0.16], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd59f193b4a48f96_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.34, 0.63, 0.01], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b80bbe3682628153_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ac92e0b890b6ccc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.13, 0.18, 0.48, 0.11, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22fef1565b8bb896_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.17, 0.06, 0.41, 0.17, 0.08], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a543b2067975acc_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_103691a9d398a011_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5afc55e0c8ecdf55_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ddc278c3c0f1ddfa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.56, 0.15], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a1528d845079e70_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.11, 0.73, 0.12], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4286f67cf12195e_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08bd7cfe88131a12_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.3, 0.04, 0.57], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7bd529620ea014f_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_056146525288873f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.38, 0.4, 0.11, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b80bbe3682628153_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.13, 0.7, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1acb8a698b381b62_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.37, 0.59, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d7148e693710227_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63f4a6135a5ea4ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5684d44875c68eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.13, 0.46, 0.27, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72fe386981e43744_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c67ebfeff6c3464d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.11, 0.75, 0.13], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05fe3bf52e6d1e57_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.35, 0.19, 0.42], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6dbcdd59e70bef3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.4, 0.42, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1967ef108e66fccc_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d48d9744aede6704_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.4, 0.4, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d7e4f38c531b1c6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.03, 0.86, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c9b91d17bed563c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.21, 0.6, 0.08], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84bd54b888ba198f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.04, 0.08, 0.36, 0.4, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18f06aa2a7033b90_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e60b934ebfe6cab8_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f2da9288a69331e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.02, 0.06, 0.9], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18a0aa3894874831_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.32, 0.47, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2395188163696e25_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.48, 0.34, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db452668cf8f9458_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.29, 0.11, 0.55, 0.05], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dced8169df28a884_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b5872bc773183a4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.96, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b4ce1c46bc3adb2_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86ed3d98659418b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c0585fc4e484a53_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cee2e5e9c7f165b_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d58df6bdbc8f988b_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1f3182d1df599c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.45, 0.37, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a41106a6764702_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_916b8cb2f81634d0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.73, 0.14], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3984b2c495c99fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa82e22dd8820be6_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef7879ac3e394e08_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65be83247834569b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.38, 0.58, 0.01], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aed0b63e5c03d329_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.04, 0.21, 0.62, 0.05], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cd8910b3526d8e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.58, 0.27, 0.01], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8869fd2b5750e391_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a104cb0d14b062fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.07, 0.31, 0.49, 0.1, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_776df851deb154e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b655876bb4420c1b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.06, 0.56, 0.26], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_726fe42b823fa468_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.69, 0.24, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05c1bbd94a26ed38_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.66, 0.16, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aed1c184fa07bb3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9338336816543274_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.84, 0.14, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72580fd4b4543676_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.73, 0.26, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75eeb59aca1b2c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.17, 0.48, 0.29, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942106c6de437eab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.07, 0.17, 0.53, 0.15, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d6b667a8d0e2e6_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01412423de2263aa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0505, 0.3838, 0.5657, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3929c4c582d3edc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.31, 0.6, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c50ff7b399fb09e7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.58, 0.42, 0.0], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80b2a5bb76171bd3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.32, 0.65, 0.02], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3184e6a0424a5d9f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.38, 0.05, 0.28, 0.08, 0.06], "state": "A task decomposition ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faa9004dd2c31b8a_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0dc08387db13270c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.21, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80aa4a56a58b0553_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71b8342812e437d6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.2, 0.12, 0.62], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c14fa2e5358e4a8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.48, 0.1], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12092eb20a316936_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f3c1d7e353fdad7_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_200b82b4723e8185_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.95, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e65619ee42fdddf_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c8317703b39efe9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.27, 0.63, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d98be830004d47cf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.47, 0.08], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fa22b6a725a70a0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.3, 0.6, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d10870c3ba3ec1e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.56, 0.22, 0.19], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c28c38584534c65_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.27, 0.7, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_446f3dbbadc0cf1a_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b0595540c20cea2_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f138cba703612e1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.05, 0.86, 0.09], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56f99c8bb8de7d43_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.8, 0.11, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4410379b943e876_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.3, 0.56, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99bab976b8a2fd83_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a8b34d4c27413eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a10129fefd69e8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.55, 0.31, 0.12], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d8f3d85bd0ed49_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.28, 0.55, 0.07], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_484c6c5fd7f02d10_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.0, 0.09, 0.76], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2e4be0ebf455f30_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.3, 0.51, 0.08], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1967ef108e66fccc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.06, 0.55, 0.34, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3094dd51c925df54_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.48, 0.08], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06b66891b017e435_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.28, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bdfa35f8de0b794_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.18, 0.43, 0.19, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d5e5fb26f04e12e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.25, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fec0d5aafb2f538_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.1, 0.53, 0.32, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3aca263e8d939555_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d00b08798b0de41_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba334f417b47b8f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24edde6920904b2c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.09, 0.59, 0.24], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_047f4022d0ade92b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.73, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3765236dc0454c22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.22, 0.65, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ad7cd62b20d0dc4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.41, 0.43, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55abc08beec17438_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.16, 0.16, 0.43, 0.1, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6beeb30b00b1f0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.58, 0.2, 0.19, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a60d7f77eaab32bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1659a620a5358cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.11, 0.38, 0.3, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_654a98c3602d6822_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_135f3b0f5b9a210b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.06, 0.32, 0.5, 0.11, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3f61057825835aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b44c56d2f74d0dd7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.17, 0.32, 0.41, 0.07, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a1d8e68b8596943_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.13, 0.43, 0.29, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee4559b273e3531f_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45dc989a35214dd9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.01, 0.97], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9200d8d7b4fd5352_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.47, 0.47, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c2a597a92a70973_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0fb96a2ce4e9d94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.36, 0.11, 0.3, 0.06, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_543113fbaf43e0c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.14, 0.45, 0.27, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c350bea5065775e4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.83, 0.01, 0.13, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd62be36cfb554f8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.29, 0.62, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70d6697f81b25fc0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.08, 0.01, 0.62], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5d94b503fa6a186_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_795285a6a5af3871_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.28, 0.15, 0.05, 0.52], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2f38149cc6f6713_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.66, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12f390717a0585ac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.74, 0.1, 0.15], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7db077d81572e34e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.37, 0.06, 0.35, 0.07, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c9987e9a8d5c892_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.93, 0.06], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8886147699a1f3ef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.23, 0.72, 0.04], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c4c29b0ba9cd2a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75c15ec4f8364337_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_748d115ea9d44aef_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d84206d71af939_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.33, 0.04, 0.45], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02dcdc11955fd5a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c083cc41faadb72f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.46, 0.43, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abe5736ee767076d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.17, 0.5, 0.26, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49ebbcfc01d4a841_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.3, 0.58, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b99e799868e6991_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.33, 0.35, 0.03, 0.29], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9412a09155e578bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16949c5f70f32545_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.3, 0.68, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df056231a7661c9e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.69, 0.06], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_076763544391681d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.11, 0.65, 0.15], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73d24f5f3af5bd78_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36aef6185ba7547a_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c88b03890da55bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.22, 0.32, 0.05, 0.41], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea787f435ab126cc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3, 0.22, 0.08, 0.4], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d36cc6e958c1b3f3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.21, 0.67, 0.1, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d94f6000e5bfb82_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.13, 0.21, 0.47, 0.08, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_720ec3c4e38f81df_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d07f61f4f041dc1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.48, 0.24, 0.22], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57f545222a6e699c_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62525facaf6e981a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.78, 0.19], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd8c8849e04681ff_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff2c14058f2a4699_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.19, 0.49, 0.26, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fbdd6cd272e07b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba20009e5f08de00_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5614d47848817c0f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.06, 0.11, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97dc77444b285ff0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.64, 0.22, 0.09], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c61f473f4d7dfa4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.22, 0.54, 0.16, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d018716c83bc711e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.7, 0.02, 0.23], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e6aa0f03e64550f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.21, 0.64, 0.06], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3169b463c94a8358_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e721bdaacd1596_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.62, 0.02], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18f9cdb9edf7a478_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.05, 0.45, 0.46, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f6ccca83dcdf267_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6a13b869121a77_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.35, 0.29, 0.17, 0.19], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98db039c333707d0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.26, 0.57, 0.07], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b0595540c20cea2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.18, 0.18], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efc7522ddc465e6d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.0, 0.98], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f4053ce0a46581e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.19, 0.21, 0.4, 0.1, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3bf06088631c40b_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f352310972059e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855831e0df346ecf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0808, 0.0707, 0.1616, 0.5354, 0.1414, 0.0101], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fadc09bc1f35ab8_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04b32369901a7a1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1bae8ddaf9d153e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.11, 0.2, 0.41, 0.19, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21e6e4c0e6933e23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.18, 0.48, 0.24, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27a1c7ccf561c523_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d671c603acb08701_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec15e95648c8753d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.32, 0.15, 0.34], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fd2ef68adb2e6b4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.47, 0.51, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aba648158aa1a6a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.08, 0.76, 0.1], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87802b8d9fdc10d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.54, 0.15, 0.12], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9caf04a72ea8393_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1841f3b777c97023_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.06, 0.36, 0.48], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5209a026db1a622b_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_188401eb16308744_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.77, 0.19, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbc2ff581eb4b2e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_783b3030434759c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8b967f8aeed4929_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.15, 0.71, 0.1], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8b9388062d1208b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.25, 0.22, 0.45], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e4c3d3e24670484_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a95602dcc4285f68_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_435f2a00676a423c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.36, 0.11, 0.31, 0.08, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52fd070b98f08e6a_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44c3a3df716fd00b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.89, 0.0, 0.03, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe211f1f0dab214f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.17, 0.65, 0.1], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee0b0410493e1cc4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.06, 0.55, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a08ba0df760fa9f4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.4, 0.09, 0.4], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08bd68d7df82aed0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.12, 0.41, 0.33, 0.05], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2ac8c87d695ed40_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_225e660bc56a39f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.48, 0.33, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d91fb78fc504c763_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.68, 0.08], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0bdeae7e396903c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.15, 0.18, 0.39, 0.11, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5391cb894b202033_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edce78236cdd2fed_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.56, 0.01, 0.37, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_516eb18cb3238cae_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.02, 0.97], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd1f046209a7933b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.15, 0.61, 0.17], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_419562476eb39fd4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.24, 0.17, 0.31, 0.11, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_554dc6f0531144bf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.14, 0.15], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97ffbc1ddf4856ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.15, 0.49, 0.3, 0.01], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_000317cb5bc25cb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8dec5643dba879bd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.04, 0.94], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31f3f664518c3633_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.37, 0.55, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1557e3813706c2bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dc5858018cb6157_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.27, 0.58, 0.07], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_321487005fce5a72_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2169f6bae8231df9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.23, 0.61, 0.08], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8f3681f6c1193b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f745051580a558be_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.14, 0.46, 0.36, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19b8d361d6c58bb5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.73, 0.24, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c9da896b7c32fa9_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c5218b87ea0fb89_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4327b4dda1b52600_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7a59cdb0515a9e7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.13, 0.07, 0.78], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cc2853d8b6012d5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.56, 0.26, 0.1], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfd8eaabd41a359f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.27, 0.41, 0.14, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d28771e0a211773d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.04, 0.15, 0.58, 0.16], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a69bf16c7172a5d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.15, 0.44, 0.29, 0.02], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66a7767c0bff758_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45dc02802258bc86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.21, 0.67, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1872980fefc17814_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aff33216ea9dd85a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.13, 0.45, 0.36, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4b6fbd9ee92d17b_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d671c603acb08701_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.31, 0.19, 0.31, 0.1, 0.04], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3f7d0e8e4ab1e20_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.58, 0.06], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c0585fc4e484a53_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0101, 0.0505, 0.3434, 0.5455, 0.0404], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bf312ca3616442b_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1a262bc94cd1453_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.38, 0.33], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc3026ef0f2905cb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.11, 0.08, 0.3, 0.29, 0.1], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8970eed6a1bafe0f_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74ec8b81ed2830f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.2, 0.74, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f948e715066f661_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.17, 0.16, 0.38, 0.17, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e99fb5d682e74a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.16, 0.5, 0.32, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afc380fc182a620b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.04, 0.26, 0.66, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62db3e8565b45145_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.35, 0.63, 0.01], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_406dc854d781f9d3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.9, 0.09, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ba5790ddedb323e_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89783f6a5bb585c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.3, 0.14, 0.36, 0.06, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_617188c23d5bf8b7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.48, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77638733736b6010_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ac47a72328dc496_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.73, 0.03, 0.07, 0.17], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_30b83d5595603d9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75059ee885d50547_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.37, 0.55, 0.02], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_699fa236d4906e01_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.9, 0.07, 0.01], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cccf80115099d7be_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbb3ff9c5571375d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.34, 0.05], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a554dfbe81015a5c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.27, 0.52, 0.07], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c2891f269fcd50e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.21, 0.69, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9388b43b4d56d5f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.05, 0.38, 0.49, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_558cb81d92c77671_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0d6226beaa81675_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.27, 0.57, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71ffb9b7352d17f7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.11, 0.7, 0.16], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7308bf3f0058c13c_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dc73f25b24fccf3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.08, 0.89], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f79ad6dc79db39d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_677bbc368d055a9a_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed19ced9352ed0a3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.38, 0.09, 0.34], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7974d1100d8f8663_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8abbeb68bc995b86_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f088c691a359dbd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.15, 0.17, 0.41, 0.14, 0.04], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9873a983b7c85a74_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd441ac060bc3d0d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.68, 0.22, 0.08], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_206192a70f30ffc2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.01, 0.93], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0650ef095ae2dd24_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca2037718b2c525f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.61, 0.17, 0.2], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e387340aafa3246d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.07, 0.67, 0.23], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bce82894511a1662_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f5d81b477029272_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ef921b6e17feaeb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.59, 0.05], "state": "A proposed workflow assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fec0d5aafb2f538_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.1, 0.85], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_179afc8b13e9b3fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.14, 0.42, 0.4, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c80d28fe994323fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea021e551d8aa539_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.11, 0.6, 0.24], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5243097fe9e634d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6335985efefdf4ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9915302d889050d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.04, 0.93], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f812f5ca76c85e4f_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e4fc6ddb6691329_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbd7380eb16fc26f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.07, 0.18, 0.62, 0.1, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acb3c2921c7a242c_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6bb37e6e1efa352_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.16, 0.34, 0.4, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_709d241c6cc26a01_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc8fee43dcbd7453_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.88, 0.11, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_797ad98a05e1e0c9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.23, 0.14], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d892f2b26a15a333_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.28, 0.64, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e8b81ecaae3da5a_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93ebf4b4dfb23aa8_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f13aa2e50a2d2bbf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.28, 0.42, 0.11, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80025f09c0bc07aa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0202, 0.0, 0.0404, 0.9394], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5561e740ebf0cd4_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c4de750e43ec9cde_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61261446916ba862_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.27, 0.09], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2517c7bfde56b254_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60e9c0bb28a3793c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.55, 0.3, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d07ac07f4b544a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.49, 0.33, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5576a093ebcc7b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.46, 0.35, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df63b4ba3d4332d2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.07, 0.9], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fdd732093a8e4e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10ad3427799710e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.07, 0.54, 0.34], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_111f2f1ce23ab7c5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.1, 0.45, 0.39, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05ee230f1e4efada_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a970709beb323a58_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.49, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb2f8bf739786db9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.31, 0.6, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c80035513f429bbd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.23, 0.05, 0.42], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70e51c10384431b5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.24, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17936cebe37839cc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.73, 0.19, 0.06], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cdf27e67ab786c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04f8e4c0aa4dbdb7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0202, 0.0606, 0.2727, 0.5455, 0.0606], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5235c825e7da5452_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.23, 0.41, 0.22, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57f545222a6e699c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.35, 0.07], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75913b977641debb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.46, 0.01], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b2f47be80766e5c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.34, 0.62, 0.02], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edce78236cdd2fed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.52, 0.34, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b01daeff4baa3d34_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01355a1bbe3ed442_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.11, 0.4, 0.36, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_573828a2a0359a96_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2242e6693ae6033e_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75c15ec4f8364337_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.6, 0.27, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdd1b49bc922fd76_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.28, 0.06], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a948ed37dd97540b_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7123dd1e6dd40212_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.2, 0.65, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_558cb81d92c77671_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.37, 0.45, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bb40e0df477319e_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45c318e75134c695_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.1, 0.17, 0.48, 0.14, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e28a5330afe70e58_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87adb7078d351e49_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.2, 0.49, 0.17, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcbb903665e0d8a1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.59, 0.36, 0.01], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e9f440ddadb1bb1_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8257a93d1597b331_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c98aafcc5e4fb11a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdcd58db49a01393_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.01, 0.01, 0.09, 0.61, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0465a90fa43140c2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.34, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40ef1e728335f865_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.27, 0.64, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c2e640271afe601_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.36, 0.52, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96ec3a4040407cf5_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d933338793dcf62f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.23, 0.39, 0.36, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06e9b1d429838007_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.27, 0.61, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21ef33cbaf481fc8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.38, 0.51, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37eb72145386aaac_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.32, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3597e69e3d0fcb37_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.31, 0.68, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42b2e86640c7c4ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c979871eafd72afb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.19, 0.47, 0.18, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ca91996a5bcc4e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.16, 0.29], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1814f6cf5d55531_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.59, 0.32, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d94256da9255c18_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.55, 0.34, 0.06, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df549ec52f99e8b5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.26, 0.45, 0.2, 0.01], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16fd45c9d5e925a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.21, 0.5, 0.22, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e65619ee42fdddf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.03, 0.09, 0.21, 0.43, 0.13], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6b8f04c6d95bc6e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.23, 0.14, 0.41, 0.09, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca3605b67eb614aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.47, 0.39, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdbeba76b5f2de50_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c1e123c10a72a48_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.79, 0.11, 0.01, 0.09], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14816238183789de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.2, 0.67, 0.09], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45ab0e3ad2727f59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.38, 0.18, 0.28, 0.08, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d8075e8bc272c29_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db452668cf8f9458_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd18e184fc1bc1d9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.31, 0.21, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_332242cd51ed0675_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9507f4ee6dab89e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eba09f30847e5e78_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.15, 0.34, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99190a8116f98039_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.13, 0.45, 0.28, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b45a658e0ec00ad1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.68, 0.12], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b76d82a44b10201_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_421aa151f47e2594_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d03f29a753f8f9ea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.18, 0.4, 0.21, 0.04], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e1018185e1de50e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.45, 0.14, 0.35], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a06245b6ec7b00_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7bd529620ea014f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.15, 0.35, 0.49, 0.01], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3ae8d9042fd2447_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.07, 0.45, 0.43, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca92046694a00f9a_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90b30e4527c34f8e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.17, 0.72, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af4dd96e411d8c26_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.06, 0.27, 0.45, 0.2, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5f92bbf2bd41b14_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.51, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_246e578804f9a3e0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.61, 0.28, 0.1], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea970192e7f14d83_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.49, 0.46, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_828cdeffa38dce37_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0dfb4a26c86e8b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.24, 0.19, 0.38, 0.08, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_487fec480332eedd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.76, 0.21, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97e27660e51feeec_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c17bc3726cb08408_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.44, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a0e33f68fb35503_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_480095c0435e9bf2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.22, 0.66, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eef4538f87ee3734_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_743d50a4d742638c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.22, 0.66, 0.04], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca18614b459f3e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.39, 0.47, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ed6d573c15094bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e074c6b049316741_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c3eaf46f60b44ed_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.26, 0.51, 0.22], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de3c3ab8a593f8d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.39, 0.45, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37f72e91ddf9eddc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0101, 0.3838, 0.5657, 0.0303], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e69099f955dc05b0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.11, 0.3, 0.5, 0.07, 0.0], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f70816b47bc2ba9_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc859c72bde181a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18428a8347470f38_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f30a84c0990011e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9039aef5521f6beb_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8933a7d560cbc9e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.07, 0.14, 0.38, 0.25, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bda9e3e06c10134b_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f9767d0d60dbb79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.45, 0.01], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db8c9047f82fa366_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_246e578804f9a3e0_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dcaf766b33254f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.05, 0.8, 0.14], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a13adf2b2684bb8c_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a77402c521accad_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1e62540d9d699f5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.11, 0.4, 0.32, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21543e68a54acece_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.12, 0.67, 0.18], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d18938c7de5feaca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.06, 0.16, 0.73], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f725e4d9636f3c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.48, 0.38, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cae845aabed7648_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9ef6c4e60169474_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.35, 0.01], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df31e953f25a8415_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5ec3dd7f102404a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.86, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ffd8aef1ff27eb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_226bec6212fdabed_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.18, 0.7, 0.12], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94824439d412b590_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e982e1d61c1a431_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.69, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2aa81e39eb618cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.12, 0.3, 0.41, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01ce05d760c3dab5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.37, 0.54, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97a160edf1344a12_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.8, 0.01, 0.03, 0.16], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c86c47d63aedd3e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.1, 0.47, 0.31, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e64cf2928acde53c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.42, 0.47, 0.1], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef7879ac3e394e08_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.19, 0.22, 0.14, 0.31, 0.11, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6000742ec2c9fb04_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa5f2d01df0b2d14_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3d928547602b222_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.39, 0.2, 0.39], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_589f7011d258c4f0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.66, 0.23, 0.09], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_514726ecff9f2185_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0113d0323359c538_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1ae8c481e775f7b_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49ebbcfc01d4a841_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.14, 0.58, 0.2], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90fb8ac3fa895a3f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.23, 0.5, 0.24, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4cfbe1e7693abbd_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23208506cd090c79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0303, 0.1414, 0.5253, 0.2727, 0.0101], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fd2ef68adb2e6b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f4053ce0a46581e_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44453df6927eec56_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3684caf19a02cec4_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caefc75f9f492970_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b55f23bf3fd4508a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.44, 0.02], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7045e0903742867f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.13, 0.49, 0.31, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1061dfbc751d1fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.57, 0.02], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dc73f25b24fccf3_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ebcd9e69f04cde8_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ffd8aef1ff27eb6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.68, 0.28, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81b291c69367ca81_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.06, 0.35, 0.46, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a57e2d0aff81f2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a67953a7b5447c6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.03, 0.67, 0.25], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25a498a7bc050295_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.57, 0.32, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b855a855ec78f96a_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6c5d1498f6c4a4c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.43, 0.31, 0.08, 0.18], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab73448bd1f6f96a_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2f2ccec9f4eb4f3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.03, 0.07, 0.45, 0.34, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7974d1100d8f8663_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.38, 0.47, 0.1], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17abb1ede5c0ddaa_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c28c38584534c65_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14c387dad35e6ea5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.15, 0.18, 0.41, 0.11, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6de91eb756d4eaee_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_381f7ed587490a54_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.48, 0.29, 0.05, 0.18], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c40a789196868c88_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d339320f715d6360_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.28, 0.13, 0.56], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce43627122618cdc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.32, 0.29, 0.05, 0.34], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd0441e3b09077c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.26, 0.14, 0.37, 0.1, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_776df851deb154e7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.54, 0.42, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dced8169df28a884_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.41, 0.46, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ceb6527e146a8d98_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.31, 0.57, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_959b4bb761f9fd3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6b8f04c6d95bc6e_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e94df78dcaf13dac_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.54, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f621483a700a1399_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.83, 0.08, 0.08], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3b5418192edda8a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.17, 0.69, 0.11], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbb3ff9c5571375d_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3377480ece924f4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.0, 0.08, 0.87], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69a37af46c1a8e4b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0404, 0.3434, 0.5354, 0.0808, 0.0], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe699100100827f9_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b8da37c3cb5eb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.25, 0.19, 0.35, 0.13, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61d185c5e2b23d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.06, 0.13, 0.39, 0.27, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad22aad8c95dd3d4_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d06fd2a0509d26f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.21, 0.66, 0.06], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_818e6a5912f415bd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.0, 0.06, 0.86], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc3b2faa809b7b4f_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_618dbd980e02fab5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.15, 0.41, 0.28, 0.06], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_417ae5dc0115cf4b_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7589675176022f36_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cec3df879b9b1b73_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.18, 0.7, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_151a6092be54c88b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.55, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4327b4dda1b52600_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.01, 0.11, 0.69, 0.12], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5bc880b1acd2086b_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89c00a58a35cc70e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.7, 0.13, 0.03, 0.14], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d3b3a97122124e3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.41, 0.01], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43b93adcf8784411_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.07, 0.9], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f868d122bbc08b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1715ee6cf1bebd6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26dbbeb886f3f1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b679fa9a6ca90b72_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52c4a4c0ef149500_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.58, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_610e3861aa143625_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.28, 0.33, 0.36], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e23d3ed5c4fb65f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.12, 0.46, 0.32, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6dbcdd59e70bef3_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bcd316b7e63330a_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a970709beb323a58_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.18, 0.74, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b3a547817eb4e73_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.52, 0.08, 0.19, 0.21], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ad49bfd98f63290_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.1, 0.87], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f0c06f473f05483_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.03, 0.06, 0.3, 0.48, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77638733736b6010_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.68, 0.21, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_417ae5dc0115cf4b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.03, 0.23, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75ba0b4562a1c07c_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50d5d63319fb5fa5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.17, 0.76, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9051a131e1e9333b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.85, 0.09, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51ab43385f943ace_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.16, 0.48, 0.18, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56719963804128a6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.56, 0.09, 0.31], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7e5c90b588d237f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.47, 0.45, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34525ee4851c3efb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.19, 0.23, 0.41, 0.06, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf8c654fa856e430_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.18, 0.5, 0.28, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0df22034803cf761_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.35, 0.1], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bdfa35f8de0b794_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.21, 0.09, 0.62], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c17bc3726cb08408_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a14af272649ab6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8182, 0.1717, 0.0101], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a5e77e78fe274b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.27, 0.67, 0.04], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88313f9c874fa8b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.48, 0.38, 0.02], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89783f6a5bb585c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e41320686c776c89_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.15, 0.63, 0.15], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70f55cf635c50299_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.66, 0.19, 0.13], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0be9b0ea271c04ea_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.23, 0.23, 0.51], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c9f7b4481634fa6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.43, 0.16, 0.39], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dda10c06e3758286_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.16, 0.08], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b679fa9a6ca90b72_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.06, 0.35, 0.44, 0.14, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d2a0010ae77930e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.41, 0.11, 0.43], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d86201a38ee571d2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.12, 0.69, 0.19], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7d9ca5f042e8a9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd4f16a66c178d22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.39, 0.43, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_786a51f0cd17b110_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.36, 0.01], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b73a2f1b5608ff5d_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba75ca2ac4dcba72_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.64, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c53d243a75d1d3d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.08, 0.63, 0.23], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72b2f1099320a361_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.47, 0.31, 0.06, 0.16], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bda017a21524e3db_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.54, 0.09, 0.25], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a7b125275824fb2_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_800c2e193528ce37_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e62d031a17af2be9_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd96ba7d24fb6302_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.19, 0.8, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb2b3a65efad6c0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.06, 0.79, 0.14], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21acd689e4f191da_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.14, 0.24, 0.4, 0.16, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b4dacde4eba41b6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.73, 0.24, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0d6fe8cfd881e51_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.47, 0.17, 0.22, 0.03, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4099269cc5fa946_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f10e2e0428d5021_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.38, 0.47, 0.11], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b4041529153c34a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.39, 0.51, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0231052b5cedaf4a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.39, 0.57], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_951d66c254d57598_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.07, 0.7, 0.2], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5860f25b7abcede6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.25, 0.65, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcc791ba43cc82d1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.25, 0.07, 0.16, 0.52], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b78a327f2b930281_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b1b0fa116a38301_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65be83247834569b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.04, 0.54, 0.39], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_435f2a00676a423c_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11fa02668357f60c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.5, 0.32, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8db604bb4e97e3c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.14, 0.46, 0.27, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b113691808a2ec60_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.49, 0.47, 0.01], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_864505e3705ebc96_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.44, 0.47, 0.08], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05fe3bf52e6d1e57_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13c8abd3e6749b00_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_119a4781b9e3fa2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.16, 0.48, 0.19, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5da00d450aa4f30_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.06, 0.13, 0.07], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bad182e4638262c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc99d808a355c871_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.03, 0.54, 0.41], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0be927482a82c71_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.11, 0.07], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_609ff41c5b4fe96e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.31, 0.09, 0.42], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b04a19e77cb02ef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.25, 0.43, 0.3, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5069f4b072635805_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.03, 0.05, 0.24, 0.5, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_844c4b85d09c088e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6253232b93fd5f52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.09, 0.16, 0.42, 0.14, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ab7589464b00097_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc76d15a34d3fcc1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.03, 0.18, 0.62, 0.09], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e5b50b9358a2ac1_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c810572ac7ab5fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.39, 0.51, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a71192f72a0dc26_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.58, 0.38, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_948cb01e11eca4cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b107e2fd2810e336_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee4ae26bf8284342_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.34, 0.09, 0.26], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9dbba3b96ecb1d2b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.48, 0.2], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c9e90d4ff10c448_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.48, 0.12, 0.04, 0.36], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_654a2f51e695bf13_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.98, 0.01, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1841f3b777c97023_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.16, 0.38, 0.27, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8dad925a0816bf17_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.83, 0.14, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56d1c8335e745f07_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfe5f723623953f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.1, 0.25, 0.16], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_722ebc0a44b93a23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.22, 0.66, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b04d58054fc1a12_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75e18403c29c782_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.11, 0.21, 0.53, 0.12, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c7fc57e9d354d06_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75913b977641debb_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d652fe35fc8fb541_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.55, 0.01, 0.43, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7cfb391d70e902f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.79, 0.08, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6af9e9709bc33d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.06, 0.27, 0.55, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad7e39855d62418e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.03, 0.08, 0.81], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49cce07ff942abd2_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67251e7c40941068_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff4de9a138cd1a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58b83624b3f43211_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.7, 0.17, 0.1], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_baa5756b3e67a276_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e28a5330afe70e58_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.26, 0.31, 0.4], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fbfb624f7c1325c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.66, 0.24, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c0585fc4e484a53_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.22, 0.28, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23abde15345cdf9a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.19, 0.21, 0.43, 0.07, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19b8d361d6c58bb5_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b957f2f047e50c4a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.28, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e97e2ad8dfc9749_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4de88fab232d27f5_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66102e2ed19b35f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.19, 0.51, 0.28, 0.01], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e074c6b049316741_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.12, 0.73, 0.12], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e8f99d6b0cbcc90_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.44, 0.37, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0143d414f0d3d86a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7b903b6266888c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.09, 0.16, 0.48, 0.14, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7bfc494df59b5cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.33, 0.5, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44719e17d468f162_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.36, 0.08, 0.55], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63c5fb84521bb8a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.31, 0.6, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3efd336e433c2aa2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.15, 0.23, 0.41, 0.14, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc66b9c6410fe1c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.64, 0.18, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ad07669cfd1dd7e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.44, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1109e15f85251d89_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.48, 0.49, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b46507de75ae98eb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.44, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37eb9a0fe1166b88_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.22, 0.73], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44d5b4bc290ef385_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.05, 0.32, 0.53, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93c30bef3767a53c_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37795e7c34811c1c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.13, 0.78, 0.09], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad2e3553df598ce4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.19, 0.56, 0.22, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea2ead956484abce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.17, 0.72, 0.08], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e866fb21121eb12_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1b2f97f93a4a614_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.16, 0.19, 0.39, 0.14, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c995be57ffee36e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc8fee43dcbd7453_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.16, 0.42, 0.33, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_206192a70f30ffc2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.13, 0.8, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_149225c65f9affd1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.49, 0.27, 0.21], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f170fe50b0271eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0422fe0c4a2a954e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.07, 0.68, 0.21], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac6285784d4a38c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.05, 0.28, 0.54, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b77cc6e07a66527_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee1f2666cdb94ace_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_959b4bb761f9fd3e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.08, 0.21, 0.49, 0.18, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea6d0c75c9d7f367_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.22, 0.53, 0.14], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bf784fbd3d66fa0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.19, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8bdde13c06e1163_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.24, 0.75, 0.0], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f15d697611568c3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.04, 0.86, 0.09], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1896a282d360f25a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.83, 0.15, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_218e11ba4e8a609d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.23, 0.63, 0.07], "state": "An agent plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19cc9ce0d60170cb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.14, 0.69, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3d5bdbb2235ec94_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f64aad792c95d5d4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.24, 0.62, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfd88908d0ac28ff_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.46, 0.05], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2605f16e337f331_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1411be854476750_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6ba799d568c4552_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.32, 0.57, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63baae22fbb14cbb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.26, 0.19, 0.15, 0.4], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad4ddb9c5f163744_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.06, 0.46, 0.39, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58fb3ead4216207e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.0, 0.97], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75eeb59aca1b2c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25e3f2c6038e1149_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e80bfd1d34c54e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.06, 0.46, 0.39, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b63953dd4904fa4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.15, 0.51, 0.21, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff28c8fedfc8892_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.44, 0.18], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bec620e56256871_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06a015811e9a4fc2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.34, 0.62, 0.02], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5d8eb7c65766023_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.46, 0.12, 0.33], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49a02484f2ff2471_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.14, 0.5, 0.18, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62c978148f357cf8_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5f92bbf2bd41b14_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88d9e157ca858567_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.35, 0.18, 0.26, 0.21], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18547896f5199fd4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.51, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66eeed9d5b1c6fec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.25, 0.6, 0.07], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_428a40c235dcf7e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.12, 0.34, 0.46, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6490118a00a2be8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.04, 0.26, 0.64, 0.06], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c810572ac7ab5fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74a03ad9873cae51_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.07, 0.39, 0.41, 0.1, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7cd86fc4559a479_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.05, 0.57, 0.35], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5daf40975f0e64_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64690e178be9e68f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.67, 0.24, 0.0], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664e57c7080b02a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f5efe4c49dacc0a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.59, 0.15, 0.25], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9a28e56e8d45212_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.16, 0.77, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91730b7793ec6d78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.23, 0.14, 0.36, 0.14, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84b34d1b2e577bf2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.12, 0.43, 0.37, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b66f015a29a340f_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ddcbd1e24b59753a_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03912a56a22b3fc6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.43, 0.4, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc8c468571f915b2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.13, 0.11], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3fcb5b04b27f083_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.05, 0.32, 0.52, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5237bc2854fddbd4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.23, 0.66, 0.03], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de9e038420e5edbf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.03, 0.22, 0.63, 0.04], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f39069fb8fe8e4c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.58, 0.05], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b238f8e8ab1f8f36_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.08, 0.45, 0.37, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_791ba2bd4d6dc501_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.83, 0.13, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ababd64a24d94af4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.34, 0.58, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b4983998b40db3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.13, 0.73, 0.07], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee4ae26bf8284342_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.53, 0.23, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64690e178be9e68f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.77, 0.22, 0.01], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b4c68ebd28cd2c4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.48, 0.06], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d3b475572a29f22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.52, 0.04, 0.35], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f45ac83d4a02031_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.11, 0.76, 0.1], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ffd73d6f3643cd7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.15, 0.81, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebb7dbc52e7cb1c4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.47, 0.18, 0.3], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfa876310e2169d5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.32, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c8cc2539136eeb6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.43, 0.18, 0.02, 0.37], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a11c67c14555a9d_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5286466a7dbbbdf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.5, 0.37, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_012b138e072c65fc_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7934137e6657d08_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.11, 0.16, 0.42, 0.15, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ed6d573c15094bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.07, 0.41, 0.43, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9f1e23f7cc8ba44_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.42, 0.27, 0.21], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e42e7fc43ad716c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52c4a4c0ef149500_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.04, 0.3, 0.59, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2eb47c6115732e19_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_427892592507f7b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.17, 0.51, 0.25, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3722ed95874190a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93bc2f37bc45a28d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.58, 0.3, 0.11], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6324d0694e9e255_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.38, 0.12, 0.45, 0.05], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe57706a36feb7e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e334abc21f31583_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16949c5f70f32545_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.4, 0.47, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab5ab0b6e54a0597_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.34, 0.26, 0.33], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_031a9b6bb6d4f79c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.49, 0.45, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbbbe0432ff72e52_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d49d90ec1f220d1a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.08, 0.45, 0.21], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e187d6ff600f7e52_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bb99492f636738d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.31, 0.31, 0.29], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_654a98c3602d6822_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.06, 0.76, 0.18], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8a881063e65d1f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_776b5d10991808a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.28, 0.21, 0.34, 0.09, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2169f6bae8231df9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.28, 0.62, 0.07, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13be773c0297e86d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.38, 0.46, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c5df943cdaabba5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.13, 0.28, 0.41, 0.09, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa316cc43372027e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.4, 0.45, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b567866ab85b9b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.39, 0.52, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_873b5e2860432e39_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.63, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9de3c41efd4c166c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.56, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2dc2e07037e910c8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0101, 0.8081, 0.1414, 0.0404], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47202d825a764672_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dff3f4053739b5b_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_026cbd92e144aace_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.06, 0.5, 0.38], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed8447c104615b87_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f05892fe0a8c7480_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.45, 0.26, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b472344d7024138_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.19, 0.07], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6f3e0eaaae957d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d58a7fdc31dca408_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.51, 0.39, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8662b4fa585b027c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.24, 0.44, 0.11, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0ab3d681769ef13_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.38, 0.19, 0.38], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0960237b382aa9a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2cb7aa54e526b1b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcead17893dd7d0f_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1de33dcf75a9ad38_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21fcf6e75dc00c2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5891a2ced68dec9_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80aa4a56a58b0553_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.25, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faf7d96f42026c39_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0fe76d4c16c28ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4048346a3001a13_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.28, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5b71794debfe7dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e41fc0aee18f1c5a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.21, 0.16, 0.45, 0.08, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4743134b82746ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6340560d22234e9a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.11, 0.15, 0.45, 0.16, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_484c6c5fd7f02d10_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6db63a9b4174d3dc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.44, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd18c81a633c8160_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a151a60a891bd4f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.47, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50d5d63319fb5fa5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.28, 0.53, 0.06], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_642cb90838e29889_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.24, 0.66, 0.05], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e5708240723dd0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.5, 0.42, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34ae804128c31609_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.84, 0.12, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44719e17d468f162_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.11, 0.17, 0.43, 0.16, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958385dfbf30315c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.4, 0.37, 0.22], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a14af272649ab6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.17, 0.53, 0.2, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e071f06d236c9a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8869fd2b5750e391_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.47, 0.24, 0.26], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0fca341cc136a09_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca3d7179e438f446_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80025f09c0bc07aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de3c3ab8a593f8d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb0557ef1b036c47_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db4f2be55cc8df46_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.19, 0.7, 0.06], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffbb734c82a3d75c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.04, 0.79, 0.13], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8700d591eb5992f7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.68, 0.05, 0.18, 0.09], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fcf46e4e0ae9ba7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.39, 0.44, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e085a5b0dd0e4212_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.08, 0.74, 0.14], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b21642dd21a2c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62525facaf6e981a_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5da00d450aa4f30_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.36, 0.52, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ddbb4775aff4ab3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.73, 0.24, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5776e09ffb9c0feb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.14, 0.64, 0.14], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc5e620a1d4b512a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.55, 0.31, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b3eabf7b7235eee_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_228949039af8c595_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.15, 0.45, 0.24, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fdac7ababff7397a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.07, 0.15, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31e1bfdefc81e199_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.17, 0.73, 0.07], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5069f4b072635805_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.64, 0.15], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95d25030f2ae9e31_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.59, 0.35, 0.05], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43cb5f8a9aa25958_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.31, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8232747b952b2e8e_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f3b816c3025bcb0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.07, 0.07, 0.83], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2df20a4352a46d2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.16, 0.53, 0.23, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3feee735ee025de_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.24, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_233471807e869217_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_379442308dfedbd7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.76, 0.1], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de405fdf4136611f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e24e2f5ea3157d6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.14, 0.49, 0.25, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43d1a96eb9414e7a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.42, 0.26, 0.31], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca94f6743f0238e_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89f4302bd069fd4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.04, 0.6, 0.35, 0.01], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fadc09bc1f35ab8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.08, 0.16, 0.5, 0.21, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7053a11502fa67ad_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.83, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6002eeb5233dee24_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.26, 0.64, 0.05, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_063ff4acd09b4cfb_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25cafd0a29d9d807_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a213d978aa6d3a_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_decd5f546f09c6de_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.07, 0.86, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0995a578ee78cce_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.02, 0.79, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f615207b8ac41f0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.37, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_347861c401fb1ac1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.52, 0.38, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e326816e61cdd0a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d4fa72aa5dc730c_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10e7d1c544d628fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4513ebc3a2e4fada_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca385d03bbb8b60_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.21, 0.2, 0.38, 0.09, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df549ec52f99e8b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed4b1a372acfcfe5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.46, 0.43, 0.01], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1109e15f85251d89_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.15, 0.77, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c44c4c9474fbf9a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.1, 0.5, 0.34], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4305e5d1d3c06f01_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.63, 0.09], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02ef1aee0da9f8fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08a211b33848eb52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.93, 0.04], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f117c0f0bf56573a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.78, 0.17, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb169bccf2c60e39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.28, 0.59, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a970709beb323a58_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae10890ce4c5ed94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.08, 0.19, 0.42, 0.16, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_747e047521520a7d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.64, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b904307e7bddbf30_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.39, 0.2, 0.23], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d3450964d628f83_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a2397e9e6ca9b2c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.41, 0.13], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6340560d22234e9a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.63, 0.08, 0.17], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb87d018c2278c4b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.37, 0.5, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2307369a90ee683_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcdbd4d1f8d5e174_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_433958e5af47da38_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03c241c77628042a_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02bca45a9df9833e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0, 0.0, 0.0202, 0.404, 0.5556], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d95067d6d81c028_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_519d267c915a30ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.2, 0.75, 0.05], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf07a6e6c4959380_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.63, 0.35, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5614d47848817c0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.36, 0.57, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e7d43768b2cbf64_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.55, 0.04], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9594c6d840194460_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.41, 0.15, 0.41], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d1dd92841725d2e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.11, 0.67, 0.16], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_880905eebc7f9186_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17373deb521a8227_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.84, 0.02, 0.11, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4b3ff45d111c75e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.16, 0.5, 0.31, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98bd266dc9ba6a31_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8f4aa9759289ac0_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31a39ba952ca88e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd96ba7d24fb6302_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.38, 0.42, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f411c0500168614_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.22, 0.7, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6f99864b0419eba_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ec9e05f6d541467_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.05, 0.75, 0.17], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c3fd4760f62c226_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.65, 0.32, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a05414d8252f8ae6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.08, 0.13, 0.39, 0.27, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d6a191ab8d94287_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.08, 0.02, 0.61], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9153760f9f3ddc9c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.31, 0.5, 0.18], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8df630f1c89defd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.75, 0.17, 0.07], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fffb1a98c8f69ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac758c61b81e011a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.07, 0.28, 0.5, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664137f0f040ed34_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fdd732093a8e4e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.35, 0.02], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ababd64a24d94af4_n", "kind": "noul", "options": ["false", "true"], "target": [0.24, 0.76], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59a2d242b0a74d68_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.05, 0.54, 0.36], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2f38149cc6f6713_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_680974644b4329c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.15, 0.09, 0.43, 0.21, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b5b829bb4db5db4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.06, 0.27, 0.54, 0.07], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97eb4ec4af748bfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e109a1030ed525f_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d86201a38ee571d2_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29e20050642fc96e_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8802e3457ff96a3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.17, 0.58, 0.19, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9186f39e644d0c75_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1414, 0.2828, 0.1414, 0.3131, 0.101, 0.0202], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_609379d67c1c7604_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.43, 0.09, 0.47], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f0414f57d29acf1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.36, 0.3, 0.06, 0.28], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e5740861647e274_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.17, 0.74, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_381ea21ab0ff3545_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.26, 0.61, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f98acc8fb667ed5f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.32, 0.58, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2267478b77d38bff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.24, 0.71, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a35b447793ac5a8c_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7d53b7924953a6f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.56, 0.36, 0.07], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0188f64b18f0d5c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_096e795313afed79_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d39c1722346d1e3d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.77, 0.09, 0.12], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab944c47b3d79d16_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.17, 0.58, 0.2, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_381f7ed587490a54_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01412423de2263aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.11, 0.31, 0.41, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39919b1ff3e12e7e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.52, 0.4], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28fa653b2eb7b43c_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d92860331691a23_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.01, 0.07, 0.77], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8abeed1e39d4bbb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45ab0e3ad2727f59_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.11, 0.47, 0.16], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c2ff97b6d31a85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.13, 0.61, 0.16], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e51735c0244eae6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.36, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da7b7735c3d6707b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.59, 0.4, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ef9ea52fd57c913_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.12, 0.58, 0.23, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e23d3ed5c4fb65f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.28, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec1408e7cb758df_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02208119118eedd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1369ff7780ac5367_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.08, 0.31, 0.49, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c007c734adb75dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.43, 0.43, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27cc1bb79b15da17_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b8304f82804214e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.36, 0.16, 0.26, 0.08, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01412423de2263aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_726fe42b823fa468_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56de72d58106f290_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.12, 0.58, 0.23, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1d15a036295dbc0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.24, 0.72, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e9bf19d0ef72f20_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.1, 0.16, 0.36, 0.27, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_07040ed65fd98c14_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.33, 0.54, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65da0eb8fbba155d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.24, 0.59, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de0aea36329054c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.7, 0.15, 0.14], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca9a4252b1c46c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.06, 0.15, 0.45, 0.23, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a410b4d9290ae21_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.27, 0.05, 0.64, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f931558db8cd32ce_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c14fa2e5358e4a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.11, 0.36, 0.49, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad0a407a4bb2b1c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de9e038420e5edbf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.93, 0.03], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81e9ddf99f2a683a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.04, 0.72, 0.18], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91feaf377d0a755a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.06, 0.51, 0.34, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bb99492f636738d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.35, 0.47, 0.04], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42c3eaf0ad0be25f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.07, 0.21, 0.4, 0.27, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed98613e6f982c13_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0c127818b7718ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.28, 0.53, 0.1], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed4b1a372acfcfe5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.19, 0.1, 0.63], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73453feed06a07e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e2bea7dd5b49f11_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.15, 0.27], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3143aa8812ab15c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0101, 0.0303, 0.3434, 0.5455, 0.0202], "state": "An agent plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da08c550f2eb0000_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a9a349005728f0c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de4d647b19942576_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.7, 0.22, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6f99864b0419eba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.05, 0.75, 0.17], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aeaa72abfe290c5e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.12, 0.49, 0.3, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_056146525288873f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.78, 0.13, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ce01c961ebdf743_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.1, 0.18, 0.56, 0.06, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f14d01e75713386_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.49, 0.2, 0.26], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_630a9af49060e5ab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.33, 0.07, 0.31, 0.08, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d8a6aaa5cba476a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.06, 0.12, 0.41, 0.3, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c11fdefaeee82fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.07, 0.69, 0.22], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a064f56facd998b8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.29, 0.15, 0.33, 0.12, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855831e0df346ecf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.17, 0.65, 0.1], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc934995de493385_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.08, 0.74, 0.16], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08b9fbdc44f3716e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.2, 0.7, 0.07], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40405ee6417b3a11_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f78da8915277b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8392d13d9f2054c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.52, 0.37, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e330d12a895ef8a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.11, 0.81, 0.08], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2e22c8aa30ac982_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6193b0bc82dcfba8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.64, 0.21, 0.09], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bdd1a9c908dfdf6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.29, 0.59, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d946e6895c48574e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.11, 0.61, 0.19, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_117bddc11d753bc9_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cda5e0ac04c178a5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.59, 0.28, 0.11], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ab68c85a6b1c382_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a16603bbadfb5b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c54e23bc2145736_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_760586154e17d289_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c539d91eec364a8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.86, 0.08, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_357f8f38c41b86f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b04bdcb096a65ae_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.86, 0.11, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2803dea13d55c33_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.35, 0.08, 0.26], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9aad87276a489840_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.0, 0.01, 0.79], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_532fc42cf951190b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.25, 0.18, 0.4, 0.08, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883c2559626f4e83_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9986f2f3cfc5ca97_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9449ed737dae8b74_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.49, 0.14, 0.19, 0.05, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_427892592507f7b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2581ca94f0fec46f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.36, 0.56, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24e6ffa6bdce6207_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f566ae45ea7d0c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.26, 0.19, 0.33, 0.09, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55abc08beec17438_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.57, 0.17, 0.02, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75eeb59aca1b2c2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.72, 0.19, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c9987e9a8d5c892_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.16, 0.19, 0.43, 0.1, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5301f774082709b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.25, 0.64, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c2891f269fcd50e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.26, 0.17, 0.54], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1754ac0f9a728723_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d00f6a757531e258_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.03, 0.96], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad13746280d27622_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.04, 0.49, 0.42], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac4425155510fdae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.29, 0.15, 0.37, 0.1, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdc84e300710c66b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.14, 0.48, 0.32, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_760b68fbf64a62b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.63, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74487a1b946d1919_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93967484cf776134_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_557fbf4a872ae046_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.28, 0.33, 0.26], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44d5b4bc290ef385_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ac721995d72461c_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f10fa7f1ab733145_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.52, 0.1, 0.37, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_517bd82e24735363_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.45, 0.5, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e28f9bd6ddadb14_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a7b125275824fb2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.19, 0.7, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_179c9e5eb6d9d06a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.12, 0.47, 0.35, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_090a6a372ae4df15_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.13, 0.44, 0.29, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0995a578ee78cce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.3, 0.62, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba2b7f5e0983a194_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.65, 0.33, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958b058965c1b030_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.13, 0.59, 0.23], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d2374f68a174632_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.06, 0.91], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_573828a2a0359a96_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5764e02d4cdce2b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.37, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4465e1914854dc9_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b637ac7d0986ce6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.19, 0.05], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab798e52d496dd5e_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9e1908f31f254e5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.3, 0.11, 0.56], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a47e0b6a296bccbe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.37, 0.11, 0.02, 0.5], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f76fc6c53e5831a_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5d2c673372e2590_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f091b0967ce7ff53_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.0, 0.01, 0.09, 0.65, 0.22], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c286467b8cc106_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c2a597a92a70973_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.11, 0.83, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f17549934649ab44_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50abb6b0785e2ef1_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_504e7f1a770f7fcc_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf791c192f7127d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0909, 0.2828, 0.5556, 0.0404], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8257a93d1597b331_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.15, 0.74, 0.05], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c5e45ab8960de9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35220e8e83d2c5cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.66, 0.2], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff28c8fedfc8892_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e554f090acc8542a_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a11c67c14555a9d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.42, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc8f2bb182152175_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.81, 0.04, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f12e76a8a4df476_n", "kind": "noul", "options": ["false", "true"], "target": [0.18, 0.82], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e3c541ad8f4c6bf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.78, 0.19, 0.02], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5396adf9bbbd700e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.28, 0.62, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65d1dc0ca3bc747c_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6488c5d9f0c7ce89_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9b0e8e627a71d9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.23, 0.48, 0.14, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec1f535871b5939d_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8ce053bf9ae403f_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0378ea7dc1e4f7b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8ce88e75a3459b2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0960237b382aa9a5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.42, 0.29, 0.28], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ac85f8a4234dedf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.29, 0.52, 0.1, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0a2b29cf61d3b1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.17, 0.5, 0.26, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea1ead63b89e9259_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.2, 0.65, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56387005a8233366_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c1cd571d2707e14_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.4, 0.44, 0.02], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f102aa5eaa56132b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.52, 0.42, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6563add88809eda7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.24, 0.58, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f329d35e680482d2_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a31c898509ec7baa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.55, 0.45, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_854b8a74e03195c4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.26, 0.16, 0.55], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15213e83e77648c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.2, 0.71, 0.07], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_844c4b85d09c088e_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_435dc03e02aab303_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_738eb66f1f2166ee_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.89, 0.01, 0.07, 0.03], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a979b919d1d3034_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.1, 0.14, 0.39, 0.23, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4966af296c2b0d08_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.31, 0.5, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2094f915535298e0_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a715b907e4a223ed_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.12, 0.49, 0.39], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_316db6800b7d00c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.45, 0.35, 0.06, 0.14], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62db3e8565b45145_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.36, 0.15], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55bb6351f9613d89_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.18, 0.68, 0.08], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87802b8d9fdc10d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.17, 0.2, 0.39, 0.16, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933e1d34a769fa51_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.35, 0.55, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a308c45605247bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc840a5700c38873_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.5, 0.01], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ea4679686ae07d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.19, 0.47, 0.27, 0.02], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bdfa35f8de0b794_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e39dd521aaba9bd_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ce2adc22e3a712_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.22, 0.69, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_462058791d0c75c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.26, 0.5, 0.23], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1ed3b6f9d499ab9_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c9da896b7c32fa9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.19, 0.43, 0.31, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4761fe14f277aab3_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a2397e9e6ca9b2c_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37a84049a4fbdb8b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.25, 0.68, 0.06], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6a044dcf0377ef5_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a829c511b6f1d38b_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9594c6d840194460_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.15, 0.4, 0.26, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba24eb7453ad25eb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.76, 0.01, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a1528d845079e70_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.18, 0.79, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c5b244facb79004_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.7, 0.11, 0.1], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4495c0424baafe5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.75, 0.06], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da7b7735c3d6707b_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a064f56facd998b8_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea24c529541e4073_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d78e27fc451a3646_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.09, 0.24, 0.47, 0.17, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_301b4b8c515c1ecd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.86, 0.08], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d33b25c1d7c9dca_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f96d485ce9b29d1_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a913987772d2ede8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.34, 0.08, 0.46], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9373c87b21dc0745_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.05, 0.76, 0.18], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_852735939cefedfe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.23, 0.66, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e7d43768b2cbf64_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da943e8d70859ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.32, 0.18, 0.36, 0.05, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eb7d16210199b8c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.46, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0efb6ce86845dab_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.57, 0.12, 0.18], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8efa5db44975027_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a4a0a4e6c7e64e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.1, 0.18, 0.49, 0.17, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee1f2666cdb94ace_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.19, 0.2, 0.39, 0.1, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc3b2faa809b7b4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.43, 0.4, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d07f61f4f041dc1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.11, 0.39, 0.34, 0.04], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba75a7a8ec9bc9b7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.44, 0.51, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83aee86b399ea3bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af4aec65a448b8fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.51, 0.16, 0.07, 0.26], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c539d91eec364a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0bdeae7e396903c_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d97b70365c031da_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66102e2ed19b35f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.1, 0.66, 0.04], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aae0d29dcf2f08ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50fc9139606ff7ce_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.59, 0.33, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e82e732f93e60dc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.09, 0.88], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a10bf0f4c787697_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.62, 0.29, 0.06], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eafcfcf2d5d7db78_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c302e9295a4a516e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.86, 0.11], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efb1ba9965dd9a29_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75816a44b8267847_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_975f0198d7d360bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.67, 0.31, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea00f02b194aa3ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ee4f1b4d9c8ca45_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.36, 0.62, 0.01], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_519d267c915a30ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56387005a8233366_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0505, 0.202, 0.404, 0.2828, 0.0303], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32e1d0bd2612c231_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.42, 0.43, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dddaa76826ef1ebc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.05, 0.43, 0.41, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae591bd08c92119a_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7237c523643d12d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.34, 0.56, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5045901decfc8a78_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.86, 0.13, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49de0dbf0922af98_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d4e650fbd5c00a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e22beddea8e9179f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.14, 0.51, 0.24, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6a684036f1696c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.29, 0.13], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9dedbf61c2cafe72_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66102e2ed19b35f_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d8f3d85bd0ed49_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d36cc6e958c1b3f3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.11, 0.22, 0.48, 0.15, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_119cd2f84a07011d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.09, 0.42], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_adf131cb77662c46_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.83, 0.12, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5c84133c8316ac5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.11, 0.16, 0.49, 0.11, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52524896ae76d0dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.04, 0.08, 0.42, 0.38], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c07da8885bf89fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.68, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6988554bbd5d49c0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.19, 0.53, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f62571168dd90923_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.0, 0.99], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_096e795313afed79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.05, 0.93], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f77ea8ed855743db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.34, 0.56, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7630f97850e9cb8e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0101, 0.0202, 0.1919, 0.6869, 0.0505], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a31c898509ec7baa_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5e68fd6a2cbdb9f_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8c596727be74773_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.12, 0.15, 0.4, 0.17, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_658709dcd9250ee7_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10666fc2532c9e75_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.24, 0.55, 0.16, 0.01], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d3b475572a29f22_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e123a4fc4a6acee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.07, 0.51, 0.4, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3a7106f06421323_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.39, 0.48, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f812f5ca76c85e4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.1, 0.53, 0.26, 0.01], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f5efe4c49dacc0a_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44453df6927eec56_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.17, 0.64, 0.15, 0.01], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7589675176022f36_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.58, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457c862f01de706b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.16, 0.54, 0.12, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d7148e693710227_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.4, 0.04, 0.5], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25d050326cd9a179_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5657, 0.3939, 0.0404], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d5c8a702c24596_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faf7d96f42026c39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0303, 0.3939, 0.5556, 0.0202], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19bfcec860ee3b1b_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90d6f7529640043c_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08b9fbdc44f3716e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.16, 0.1, 0.7], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81b291c69367ca81_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883b3ca38a9fc3e7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.84, 0.03, 0.1, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2f455f8f78cafcc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.41, 0.21, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5d94b503fa6a186_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54121102ffa889d4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.09, 0.65, 0.23], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_333992169e79c9df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.3, 0.6, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d857804cdfc5c101_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.08, 0.53, 0.3, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb35d3e8c0e5c00e_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff3613b6a58ada98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.82, 0.12, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1cb6cc228707ade_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.24, 0.63, 0.05], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b11eb13b5917330_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.38, 0.52, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8074deb77843f9db_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.83, 0.14, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb76cdd8e743e2d8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.24, 0.64, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ec36d785c6015ba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.43, 0.04, 0.42], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fb36b0f54a8872d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.13, 0.36, 0.32, 0.09, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a48ac86858e07cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.55, 0.3, 0.02], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f98acc8fb667ed5f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.39, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfc11fb6b7047bad_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.75, 0.05], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9934de50556b441_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef90bf52e81314da_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.63, 0.23], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53fadb4a6aa3162d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.13, 0.19, 0.49, 0.13, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae1c2148cedc0acf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.07, 0.46, 0.11, 0.36], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bde3a7faf13bca5b_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_247bfde25b914dcc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.1, 0.67, 0.17], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3e6eed98e0adad8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.33, 0.46, 0.14, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5485a116a01bed_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0dca1c074659eed_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.52, 0.22, 0.06, 0.2], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533d3c740b37e234_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.75, 0.23, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38862ae2ba1f7e57_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.56, 0.41, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e13cb220fc6916b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54d569144b0711b1_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b21b3c3bbeb5bb9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0188f64b18f0d5c0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.78, 0.18, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6375e93f2f6ae957_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.08, 0.2, 0.49, 0.13, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1e32d70bd91fd37_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b7e7d605268071b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.8, 0.16, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df79e920cf5d75d8_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1ae8c481e775f7b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.42, 0.5, 0.02], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c46f1b6d2bf47a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.16, 0.56, 0.19, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dde8333f9126638_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f68e4963b6bca43f_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e40339b2c1ba42f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.24, 0.49, 0.15, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d59f2f3bff1cafc_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f148582b365c909_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.03, 0.07, 0.48, 0.29, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13c9b9f944881c4a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.21, 0.61, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d99145ed5d7c769_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7934137e6657d08_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.9, 0.04, 0.05], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c46f1b6d2bf47a5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.19, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_056146525288873f_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c33b0e0b39ed5537_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.19, 0.22, 0.36, 0.09, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_487fec480332eedd_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e60c406d72906ef8_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f10fa7f1ab733145_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883b3ca38a9fc3e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.24, 0.51, 0.18, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43696ef77c88edbd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.07, 0.14, 0.55, 0.16, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74a45807872fc87a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa7ff157e0adc5de_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4da478a96ee240d7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.17, 0.18, 0.32, 0.18, 0.09], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95c1a1cb91750dfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0303, 0.1616, 0.6869, 0.0909], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45dc02802258bc86_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fe400bacdd7c19c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.33, 0.58, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29fa85b639d7e3f5_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_622cd3abd9843a56_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.09, 0.14, 0.44, 0.25, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dd8ed7513e8c443_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.08, 0.63, 0.24], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2349a345668b42c3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.4, 0.02], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e43971168e23c79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.23, 0.18, 0.35, 0.07, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e704a65f6f83cfec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.46, 0.4, 0.02], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04ef04516d22c38b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.08, 0.41, 0.46, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72576661340036c3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.03, 0.87, 0.08], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f40eaca60df888e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.35, 0.51, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b904307e7bddbf30_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.11, 0.34, 0.38, 0.07], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b5d5573f3135faa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.96, 0.0], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06e9b1d429838007_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.71, 0.21, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_446f3dbbadc0cf1a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.35, 0.14, 0.43, 0.08], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_339bf2f73c6e149e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.17, 0.16, 0.41, 0.13, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6071d1f8adfbb87_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_822f019f3f6ab89f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.48, 0.32, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36b7776629c7f6fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.08, 0.33, 0.43, 0.14, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_852735939cefedfe_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31bd35cdb43705a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5069f4b072635805_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1261402c537bd438_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.17, 0.24], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39d3b654c286f635_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e55208dbc6434914_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d690ab377aab9de0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.37, 0.11], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fc0957fb28cd34f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.11, 0.42, 0.35, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9857bad442b7f52a_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47d6ee183a553824_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.45, 0.42, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dc5858018cb6157_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.92, 0.08, 0.0], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41fd94c0d9caad07_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.3, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31bd35cdb43705a1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.14, 0.15, 0.45, 0.12, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_261e972ebb0362ee_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.94, 0.05], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c8d5ad73f0e75ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.38, 0.53, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5778e094db6dfb27_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.33, 0.61, 0.02], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8292c1f469b0ad1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.0, 0.04, 0.87], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f218152faa87316_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.03, 0.11, 0.47, 0.28, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c452c5cefe89f758_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.18, 0.76, 0.04], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c00c0f6e22d6e00f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.76, 0.05], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfa876310e2169d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.44, 0.37, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d5192abc8a918e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.2, 0.47, 0.2, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d94256da9255c18_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77417bed6cf76430_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8d045508aa2974f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.29, 0.56, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75405c9a48fe73ed_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.59, 0.17, 0.03, 0.21], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f853e4d570293112_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.28, 0.05, 0.37], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9ab39462e4131ae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.13, 0.49, 0.29, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15934917b0f598e2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.62, 0.34, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_440a473d3c30c59a_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6f91126611ee44c_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_754c9e8b6d9352c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.15, 0.5, 0.25, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bc8b24a2f29e037_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.32, 0.52, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_546701b9a7d0a50c_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9507f4ee6dab89e7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.44, 0.17], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2292abe1e27d7b14_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.27, 0.65, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08a211b33848eb52_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0dc56ca0a4e15a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.43, 0.45, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_736d4923cc9210b5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.38, 0.59, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a95f3d73ec727610_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f10e2e0428d5021_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.02, 0.19, 0.64, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26165dd77cb85ae4_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83ab6e296c9f7d49_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.32, 0.6, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_843cc9bce04020fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.21, 0.7, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51bdf571be3b5933_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.03, 0.2, 0.58, 0.07], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ad920ad1837f01_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1e3611a90fa1ecd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.18, 0.68, 0.08], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_282abe460b40b1c4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.65, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d6f18318826c74c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.19, 0.18, 0.44, 0.1, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d8b4bcaad3a4417_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e5708240723dd0_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29a8c5dc914f6ad6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.12, 0.16, 0.37, 0.21, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6af9e9709bc33d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.48, 0.4, 0.11], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d97b70365c031da_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.11, 0.71, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5daf40975f0e64_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.1, 0.47, 0.37, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c93c08e5ff47243_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.26, 0.14], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a81efe2d6089cd13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.79, 0.08], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4bfbf7f9f2686e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acd9b1b222c524b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e52b22a99831d74_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf791c192f7127d_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2c721106cc5cadf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.76, 0.22, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3abd9133fc74e47a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.02, 0.58, 0.29], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fd2ef68adb2e6b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.11, 0.51, 0.29, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87e23308981e4f35_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.71, 0.25, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06e9b1d429838007_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_863aaf03c6236e8e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.32, 0.49, 0.05], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_381ea21ab0ff3545_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aeba1cb3e4ea934_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1441e1a2a739762a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.37, 0.46, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ddac1f8244e8757_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d7723161029d043_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94e076f83aee26bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.14, 0.16, 0.34, 0.23, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f36e3f3387d1bcc_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_631efd049871575d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.04, 0.22, 0.6, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2dd1df6151e9a4c5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.19, 0.77, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a7da70dbffd3501_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.25, 0.32, 0.12, 0.31], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c6542da9aae1377_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.32, 0.13, 0.34, 0.1, 0.04], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dce96efd497b643c_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d530d2fec94d0a33_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.5, 0.3, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_784acf7120218783_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.36, 0.0], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a554dfbe81015a5c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.35, 0.37, 0.25], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fc5913112177801_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.76, 0.21, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60b06e1f0ebb154f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.3, 0.57, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95f585062ae50219_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.23, 0.19, 0.43, 0.05, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b77347f4f1490083_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.42, 0.11, 0.27, 0.09, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1caddd02c110239d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.2, 0.43, 0.28, 0.03], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_067018c50b7b46cf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.75, 0.18, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ef28fe276a4fdb0_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a213d978aa6d3a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.1, 0.42, 0.37, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f1de1e310fc29a4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.63, 0.1, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bab324bde97e7e53_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef0f862e418e0fb0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.46, 0.37, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2aa81e39eb618cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39dbb8fce4ff9aea_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f53bf6e414d7c816_n", "kind": "noul", "options": ["false", "true"], "target": [0.21, 0.79], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f138cba703612e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.09, 0.76, 0.15], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2a62b47d5879b03_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e01e2d70105cb419_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bec73f6ad91c788e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.25, 0.18, 0.32, 0.1, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c07da8885bf89fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d2a0010ae77930e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.36, 0.56, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ea81dfdb02e9cd9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.32, 0.55, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27361b65af035d09_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.77, 0.05, 0.11, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f088784d513cedd2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.46, 0.26, 0.27], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_128a5de4f81774f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c37e33704ad4cac9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.33, 0.15, 0.28, 0.08, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ae16604561b07d6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.83, 0.14, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3f3e0bce9df2ae2_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9120c1cc59c1cb9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.3, 0.08, 0.36, 0.09, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0df22034803cf761_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bc87874aaaa2a25_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e982e1d61c1a431_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1411be854476750_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.9, 0.08], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8509958fab1aeff3_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f05892fe0a8c7480_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.25, 0.1, 0.38, 0.09, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ae600ace3b488cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.5, 0.34, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4dc7f9d70144ab55_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.44, 0.11, 0.24, 0.06, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c24488f1fc4edb1_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dddfae548712705_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.63, 0.11, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eee361e9a9332d0b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.15, 0.56, 0.16], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_604c43c018667b44_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.41, 0.06], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_524ced9c71154d85_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7cd86fc4559a479_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.06, 0.85, 0.09], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1d15a036295dbc0_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_510ecf4fed3932ed_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.59, 0.33, 0.07], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d138657f3f8cbe7f_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47524887f080b56c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.46, 0.07], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94824439d412b590_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.17, 0.48, 0.21, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39fdeb6d5212dcd9_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c37e33704ad4cac9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.05, 0.51, 0.41], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05863c271559edbd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.68, 0.28], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5fd6903727bb14a_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cedb517c878f3845_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.47, 0.51, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cba376fa60c1709_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.37, 0.45, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c65c03fc3bfbca6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.08, 0.44, 0.39], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b89f315163b792d9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.46, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bc03d253e2fdee5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.33, 0.38, 0.27], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df128e195120190c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.0, 0.06, 0.88], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57c1292cc02b6514_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.36, 0.32, 0.31], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4acada9f6b95ebad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.72, 0.05], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b18885ad84792403_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.41, 0.52, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4954a5186e003a23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.59, 0.16], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a34f40cb83cc6e44_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.49, 0.09], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9bffd8b4b4c318d_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cd8910b3526d8e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0efaf93756ab9f2d_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67c7fd74dbcff67e_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_031a9b6bb6d4f79c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.07, 0.7, 0.21], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11d6ac10253a0dd9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.04, 0.42, 0.09], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c46a11442eeaf769_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.39, 0.61, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5578540292ccbcb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.13, 0.66, 0.11], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9da8db47b84fe841_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56a9b45a7d9daa5f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.15, 0.56, 0.19, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c229aed5289b89cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.22, 0.65, 0.04], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e1671396a87004c_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecbc1c68f3407614_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.14, 0.24, 0.42, 0.1, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46e25b4310197415_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbd7380eb16fc26f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.0, 0.99], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5ca8198862eba52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.96, 0.0], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed98613e6f982c13_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.38, 0.11, 0.45, 0.06], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e332feadb1fcbaa_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd11ff0c5bb4dc35_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.63, 0.31, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9bdd34ecc6fc334_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.11, 0.15, 0.3, 0.31, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14383c3bf6dfbcac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.2, 0.66, 0.12, 0.0], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53a4bd46613efe83_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.88, 0.08, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62a423b47e1cb7c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f66b13a75b7317df_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.53, 0.34, 0.12], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bf4cef3db98ebdb_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d025b48904a764b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.63, 0.05, 0.16, 0.16], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f29b86096a51f068_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.44, 0.52, 0.01], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c26adcfb2d68073_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.66, 0.11], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9b19509826dc13d_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_127cee9cbff83a52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.39, 0.07, 0.34, 0.08, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cae5e8e891d19bda_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.68, 0.03, 0.18, 0.11], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90b30e4527c34f8e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.65, 0.07, 0.25, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fb2c8fc2ff88647_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.34, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d76834f14ae6d84_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.44, 0.52, 0.01], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6e41e7cb15f6447_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91730b7793ec6d78_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.26, 0.06, 0.38], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_391826797eed8cdc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_462058791d0c75c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.3, 0.65, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a01dd2aa3738a7e0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.45, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c00a252f79f34100_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.34, 0.56, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dec118dfd39d53ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_630a9af49060e5ab_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33afe2d1a622d26d_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb85437c93ff6894_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.59, 0.27, 0.1], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a445f519e025b9b1_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bb192c0ba816fa9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.42, 0.55, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d59f2f3bff1cafc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.86, 0.1, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5333fa36f9f7007_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.14, 0.51, 0.27, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73453feed06a07e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.41, 0.58, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0422fe0c4a2a954e_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a47c0376cf080712_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57b1b3ff7615dc4d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.41, 0.18], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b58857e1e41d99c0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e332feadb1fcbaa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.27, 0.14, 0.39, 0.08, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3087b083a7ca9c59_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9347e96726f4d06_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.77, 0.19, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_924ce83fe8655c2a_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfaa15f085f75efb_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2752683b0ce259f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.11, 0.54, 0.25, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32e5c4d9325d1ed3_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4096de6d789c6d69_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0101, 0.0505, 0.3434, 0.5354, 0.0505], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c1a7e63eb1d94c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fb36b0f54a8872d_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7222b0ea154d4d3c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.48, 0.31, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b03835b56b94d32_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.17, 0.54, 0.18, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6216ca71d1f2dd5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.07, 0.28, 0.53, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27a1c7ccf561c523_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.14, 0.3, 0.4, 0.09, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c57f86afbe9b0dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b50e29c4e24db95_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1cd9e71d4018ede_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.68, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1772c9ab9fae2a96_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.55, 0.24, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1561a885ffa801fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.56, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6954279c23320ab6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.55, 0.24, 0.01], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61c0d40d4c581382_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.21, 0.12, 0.43, 0.15, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49f3bb6f4fe32cd3_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7ae565c1df8e900_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e160dca304097d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.21, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9060c353303d6444_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.17, 0.57, 0.19], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f1de1e310fc29a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.08, 0.12, 0.4, 0.25, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f95fe8772f97f41c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.18, 0.81, 0.0], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9aa654dbc67977a2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.26, 0.16, 0.39, 0.06, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_222dec12a2dc24bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.42, 0.46, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c061b9c3430adba3_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b15d0993854f063_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d43940b6c9501535_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdc84e300710c66b_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e73cdd46cc5bafad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.28, 0.16, 0.49], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7298ae159b451fac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.2, 0.73, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4de88fab232d27f5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.5, 0.46, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_527603a833da6ea3_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4953b431025795ec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.33, 0.2], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9bca918a7877c21_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.31, 0.66, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6610b67c875a471_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d33b25c1d7c9dca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.32, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dc392d28da02276_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.18, 0.66, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_894961486eefcb53_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.44, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c5e45ab8960de9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.18, 0.65, 0.1], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e570fa9aa32ad948_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.29, 0.58, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e29dd15a942b6af_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.78, 0.15, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_654a2f51e695bf13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.09, 0.31, 0.55, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5011ecb83f3444_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.07, 0.15, 0.46, 0.23, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f468b1e88e881963_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.62, 0.36, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7369bfb8e3e4fdfd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.17, 0.69, 0.09], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_720ec3c4e38f81df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.12, 0.48, 0.26, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a5fe0273aade08a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.26, 0.13], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_116b28e301cdd6d2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.1, 0.89, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3313e203b53cdcfe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.25, 0.13, 0.43, 0.09, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57791e80ef540d39_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.92, 0.01, 0.02, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4048346a3001a13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.55, 0.32, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e69099f955dc05b0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.02, 0.95], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c434137389b39363_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c452c5cefe89f758_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.1, 0.71, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd7dead90f04356a_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_085eda9f8c43acf9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.05, 0.2, 0.68], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edd61b957f1466d2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.16, 0.39, 0.11, 0.34], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85259c65fde0a38d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.76, 0.16, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b7e816f3e0e6ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02208119118eedd6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.21, 0.57, 0.17, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e70d1f7cda39b5d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e2e33110b8a1a2d_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42ae43bed99bcbb6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.53, 0.14], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c86c47d63aedd3e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.33, 0.11], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8084730dedaf4734_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.61, 0.34, 0.04], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d740bf72f15a8835_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.17, 0.68, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_076763544391681d_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_106fb08e5b5d172b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.34, 0.25], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51bdf571be3b5933_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.54, 0.42], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd1ea4d7fb29b2ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.22, 0.31, 0.43, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43c4c231c0b289fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.21, 0.53, 0.19, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b2b61ab647052f4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.02, 0.96], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_317a2c18da54412d_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd1c0a90154413de_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c8a587b88d26881_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bec73f6ad91c788e_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38a9659a5f0eb094_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.3, 0.55, 0.03], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bae339b4001a7ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.08, 0.56, 0.27, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a219998908c51706_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb72a0e06056c95d_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e1671396a87004c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.51, 0.35, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1c49d95821acf91_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.15, 0.82], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05c83edcd842ad19_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83ba125d5543f087_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.34, 0.02], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff48ee53431e8bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.96, 0.0], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0537f4fa04f987_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2edbfa54a492ca18_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.51, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d002aee1a007137a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.15, 0.16, 0.4, 0.14, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8370eb43558c3ccc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.42, 0.56, 0.0], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c16ed93f98e9c0cf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.55, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ee4c2a27a1d6ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6cad2b588c6f69_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.32, 0.2], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a445f519e025b9b1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.59, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c5df943cdaabba5_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_465b0032bad586e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.22, 0.57, 0.1], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74ec8b81ed2830f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d364f6224d92d0fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c6731465262a2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae1c2148cedc0acf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.1, 0.15, 0.48, 0.17, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9738231a2e5e108e_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8b967f8aeed4929_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.16, 0.83, 0.01], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb19b614dbaa8653_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ab5847543a2f6b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.37, 0.51, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7aa71f6e258e6304_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.1, 0.8, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b46507de75ae98eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.22, 0.45, 0.29, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8efa5db44975027_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.11, 0.58, 0.24], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95f585062ae50219_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.06, 0.57, 0.27], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74d59ce47911ead2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.34, 0.55, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_699c062e3dad48c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.27, 0.43, 0.2, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df609e0542c72381_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.75, 0.03, 0.11, 0.11], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_105cf342772e1e93_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.68, 0.27, 0.02], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415b32b9c52d106f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.2, 0.65, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8509958fab1aeff3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.05, 0.66, 0.29, 0.0], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90685740014cfc0e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.4, 0.09, 0.26, 0.09, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29968b71c569ff6c_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7093b3469fdc55d_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1056fb002cbe3e62_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.08, 0.76, 0.16], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fcf880c7bde82819_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90685740014cfc0e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.63, 0.11, 0.05, 0.21], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7d30904674b2486_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c28c38584534c65_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.12, 0.47, 0.29, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a58ec7a7bfcbd56c_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4953b431025795ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.12, 0.66, 0.18], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc934995de493385_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.7, 0.18, 0.11], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bfb609893d01230_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.61, 0.11], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f352310972059e5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.04, 0.12, 0.38, 0.33, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5212930c516f071d_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0961dc21d567c8_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_828cdeffa38dce37_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.35, 0.44, 0.19], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e65619ee42fdddf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8199c5974489484e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.08, 0.6, 0.24, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2af4e72265d54d2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb6cf4e29ec47288_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.6, 0.22, 0.13], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40fc447ea5271f26_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39c23f5624930a35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.29, 0.05, 0.33, 0.14, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dddcb0dd5d8ad5f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b879ae8191c0d6d_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e31b3fd6227e051_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c96ee0c9f61362b1_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1850cf203d2597b9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.35, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21000fa3ed8bd88e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.22, 0.75, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff4a76b023bcf36_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.31, 0.56, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a117425cb1759aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3760299571e93537_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420932e343b3f53f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.07, 0.31, 0.47, 0.06], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60e9c0bb28a3793c_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8df58a3e170d4473_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.39, 0.21, 0.36, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_179c9e5eb6d9d06a_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bce8cdb655f6892f_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b4668765d268f40_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ac92e0b890b6ccc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.46, 0.15, 0.24], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44b186cbc8a235b9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.67, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54a5968f9c2d26c9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.72, 0.2], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81e95ae83319c565_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02ddcd5932212adf_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48ae61237e67d20f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.34, 0.09, 0.34, 0.06, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d34c97eed7f654d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.8, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_282abe460b40b1c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4abd0003cbf4a7bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0707, 0.1414, 0.5556, 0.1919, 0.0101], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa9f468c44113d6f_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf3f02b36924b240_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.29, 0.09], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cd29498fb8de605_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.12, 0.2], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06b66891b017e435_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d8e390b021d0cad_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2ac8c87d695ed40_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.11, 0.85, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d81ca87440a3154_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.72, 0.2, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_172c23b0449e0268_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.33, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16635cae28fefb23_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88d9e157ca858567_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.18, 0.25, 0.4, 0.07, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68a90a47f083c553_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b4c68ebd28cd2c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.32, 0.55, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2df5ce0eea7f7aec_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea87355cb7e78d5b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.2, 0.73, 0.06], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc613c6d0aaf66e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27a1c7ccf561c523_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.6, 0.31, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3efd336e433c2aa2_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0995a578ee78cce_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99f8bfca34dc59ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.68, 0.23, 0.07], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f4ac2c7e18f944f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.1, 0.56, 0.26, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41a5483da99db0fe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.87, 0.09], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2123b8a1aed0e64c_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_767c6dc95eeb6fcd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.28, 0.63, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f02eea3042987e21_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.27, 0.05], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc5dee796b7a9d6a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.14, 0.48, 0.35, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b7c88bbeaa866f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a5e762fa5c21ad9_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_250b9a909f2c78c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.13, 0.49, 0.33, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2ad5cb7a4702e42_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e9e86be4b6f5a83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.04, 0.28, 0.58, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a9761704d49df88_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.42, 0.08], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2d6d766ec7ef188_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.32, 0.07, 0.58], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ed5f86f56792730_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.41, 0.47, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa7ff157e0adc5de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.08, 0.42, 0.46, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_122f793ad020a6e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba32c0160d76d762_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.4, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_668bc2cafea8dcd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24573e50cc8e35c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.05, 0.58, 0.21], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_569000e1f3b19cf3_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b2fc6f878b8b9f5_n", "kind": "noul", "options": ["false", "true"], "target": [0.2, 0.8], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1888572dbbc2798_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.49, 0.38, 0.12], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d35e59f8edd8f5db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.53, 0.34, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2949f6caf642b23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.43, 0.41, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35041b6f6e91c569_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.24, 0.62, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ef9decb28db7831_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.2, 0.55, 0.15], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d1cb3dfc7166aa6_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf4d0d885a7a4393_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9aad87276a489840_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e22beddea8e9179f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.59, 0.29, 0.08], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_813e5c81dc7d40f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.1, 0.48, 0.31, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af0007c0abde98fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.07, 0.42, 0.42, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c8d5ad73f0e75ba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.75, 0.22, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c7a863d47ff9f12_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c449d87bab1a6f8_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b21642dd21a2c5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.18, 0.55, 0.27], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44c2397332e5ce85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.14, 0.45, 0.28, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36b7776629c7f6fe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.57, 0.21, 0.11], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48eec294d5fed903_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.71, 0.22, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c23af1d2520f75b8_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a5fe0273aade08a_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fec0d5aafb2f538_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54fdf332a225027f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.7, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6abeab9ee56a68b_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c75832a3ac148a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.17, 0.68, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bab0f0817a36018c_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f3c1d7e353fdad7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.23, 0.65, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d138657f3f8cbe7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.29, 0.1, 0.32, 0.12, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ad920ad1837f01_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.73, 0.17, 0.07], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5c5a7a8115a934c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.05, 0.15, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e343315796609820_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.5, 0.12, 0.36, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1955980e8928d16_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5efcf89cd0f623d_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0fe76d4c16c28ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.05, 0.47, 0.38, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50329125c3824eac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.28, 0.55, 0.03], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a8d9df0c7ecc172_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.19, 0.43, 0.36, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca92046694a00f9a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.11, 0.19, 0.37, 0.2, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6c3022ddfa9750a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.22, 0.23, 0.39, 0.06, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b5132c68f7d83e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.3, 0.11, 0.33, 0.12, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b9290268ccf3f58_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.12, 0.4, 0.25, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b7118f62b93def1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.47, 0.09, 0.34], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e28a5330afe70e58_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.2, 0.47, 0.19, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9dda360e0fd8f05_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.19, 0.76, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77638733736b6010_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.5, 0.38, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70b3fc788b4930ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4a0387f351cf745_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664d0708dabb0a71_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2dd4ebb8f240660_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.39, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcead17893dd7d0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.12, 0.46, 0.3, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d18938c7de5feaca_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d039f06efb7ba4b4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.35, 0.24, 0.38], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e67f41eef1c700b0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.12, 0.15], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c8317703b39efe9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81e9ddf99f2a683a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.12, 0.59, 0.16, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3944d640adf9ebe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.21, 0.16, 0.46, 0.08, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1274b5e329255a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.09, 0.38, 0.33, 0.07], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46478b26415bf170_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ba6b51a9d789ae3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.0, 0.0, 0.02, 0.31, 0.61], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45f37b88cc4aef9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.14, 0.5, 0.3, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d5192abc8a918e4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.07, 0.7, 0.2, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8e2c13db052f0a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0a285e02941a5c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.29, 0.56, 0.05], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c58da4152a5268ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8bfa2d9adab1783_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.2, 0.46, 0.25, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3641d67b48692fff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.23, 0.68, 0.05], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab40858c20e66da9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.48, 0.16, 0.31], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38a9659a5f0eb094_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01355a1bbe3ed442_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41818955047ed9ce_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61e4e06bda6333c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.32, 0.54, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db4f2be55cc8df46_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9dad05700ff379f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.23, 0.1], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d4dee1c0942c4d6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.34, 0.53, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7058def01b14c20_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.27, 0.13, 0.01, 0.59], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffd5a1454e5e09b8_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b421120af978f63_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.51, 0.4, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f29b86096a51f068_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_934430ac92e19fa2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.25, 0.62, 0.06], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea970192e7f14d83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.36, 0.53, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97cc7ba3218e88ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.03, 0.95], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f5d81b477029272_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.13, 0.34, 0.35, 0.08], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e80e032ed2ff03fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.44, 0.51, 0.01], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_106fb08e5b5d172b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.24, 0.67, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_773018795f0555c9_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11a283b71f20ba17_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d48d9744aede6704_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.13, 0.63, 0.09], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a239a2045ad73b6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f329d35e680482d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.11, 0.47, 0.32, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_926c675d258b41fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe8e23a9fc1bbeca_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba3afe1ec466d834_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0707, 0.1717, 0.5354, 0.1919, 0.0101], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_652b46a3c48d6ba5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.91, 0.07], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13b58a8f3368a0b8_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_897cf18127c0ab61_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_566c7b9721627609_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5edb579c7d7a6b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.66, 0.01], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0d6226beaa81675_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a2a800ba6d09a32_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5e68fd6a2cbdb9f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.09, 0.49, 0.27, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29d3c5e68fee195d_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4eecb6fbf12d6e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_178fcabbbb427a4f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.76, 0.16, 0.05], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11e8b6c8c6ad6012_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.38, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89626afa95dd213c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.37, 0.51, 0.03], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d57848d82b7737b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.28, 0.65, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aed0b63e5c03d329_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.48, 0.05], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10c02a9cb53dfaaa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.1, 0.78, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c79b84590c24c1d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.17, 0.19, 0.42, 0.11, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5764e02d4cdce2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.28, 0.61, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2581ca94f0fec46f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.51, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5d94b503fa6a186_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.51, 0.41, 0.04, 0.0], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c355a3f503bfe39_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.18, 0.38], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aec8bd310383019d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.05, 0.19, 0.58, 0.09], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcb6a245ed646125_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.32, 0.67, 0.0], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54c228115ecee8b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5ec3dd7f102404a_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cfcf85c28948c68_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e5085bd7582db3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.18, 0.51, 0.29, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed19ced9352ed0a3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.23, 0.18, 0.4, 0.11, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6648b26a85a46e7a_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d22f1ffc61c3652e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.26, 0.08, 0.58], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3722ed95874190a8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.4, 0.42, 0.08, 0.1], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f3715b5e85501ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.19, 0.46, 0.25, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ead24e3fa635dd7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd70b14f59d81f96_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.63, 0.1, 0.04, 0.23], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17d3501a65a8d7a2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.69, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41d9d6d888d9a6e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.25, 0.58, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_465b0032bad586e1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.37, 0.56, 0.06], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0002c468edcbd614_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ded8c953aed34c0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.53, 0.26, 0.04, 0.17], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c61c3afb9d90a2db_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.35, 0.06, 0.54, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0143d414f0d3d86a_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_544f966967942aee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.19, 0.47, 0.15, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60b48e3d20a50666_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.39, 0.03], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecac8bd4643bdc06_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.28, 0.1], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd7092a1e7f9afc6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.43, 0.07], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aff33216ea9dd85a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.57, 0.37, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d933338793dcf62f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.36, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8e65e15a7ab9c35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.74, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70e51c10384431b5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.46, 0.42, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1794d9cbd14d2f7d_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3d8d3a99c61c671_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.09, 0.17, 0.43, 0.18, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7ae8e9079367878_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.18, 0.39, 0.22, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b1743cf5490abfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecac8bd4643bdc06_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d1d48ab06b690cb_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28eb5ced6398964e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.53, 0.3, 0.16], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c40d20c8e407380a_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a49e867ad3c4b42_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.56, 0.41, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc3abd12576d7219_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.05, 0.64, 0.21], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2af4e72265d54d2b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.84, 0.08, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ced475407d1bee31_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3209b7712c92c0be_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.08, 0.21, 0.38, 0.24, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b1743cf5490abfa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.44, 0.33, 0.11, 0.12], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e20f8d372d4969b_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_deb5e26222a441e6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.13, 0.84], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_096e795313afed79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.31, 0.46, 0.15, 0.01], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a97713c1f8d10613_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.04, 0.46, 0.45, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be6bb4e90c01ef57_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16a7daaf8df5a37c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.21, 0.76], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20a969f6e9f1e409_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.85, 0.02, 0.01, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2d8b03086a5a292_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.09, 0.68, 0.18], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5770a524b2afc506_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66a4b087937edaa4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.54, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14816238183789de_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.79, 0.17, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6166e74e86e271d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.12, 0.78, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa48d936fa4241f1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.54, 0.39, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed977694b526a4c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.32, 0.21, 0.31, 0.06, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4005eb7b441c70cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.09, 0.65, 0.21], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65994e45b04af382_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.76, 0.13, 0.08], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b04a19e77cb02ef_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.03, 0.16, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d79e2a803ffc60a3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.67, 0.27, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dec118dfd39d53ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.27, 0.64, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d2a5927935b342_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.18, 0.8, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6f99864b0419eba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.85, 0.02], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ec27ee2c7c667d3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.11, 0.04, 0.82], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86a8c356e378577a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.5, 0.26, 0.21, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e721abe284679d12_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1315fb4846c685f2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.63, 0.29, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45f37b88cc4aef9b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.69, 0.19, 0.11], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43e79444e0d58405_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.7, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5ae31fa0847e941_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5be7b99a039592c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.0, 0.22, 0.72], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ec73a7d425b6710_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.43, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d5c6484b55e3ba3_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83ba125d5543f087_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.58, 0.26, 0.01], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f464fc0f27d5cb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.2, 0.46, 0.21, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_813e5c81dc7d40f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d79bdaf2c936480_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.29, 0.63, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f77ea8ed855743db_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_586530ef2ad24e6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0, 0.0, 0.0404, 0.5455, 0.3939], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_baa5756b3e67a276_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.33, 0.09, 0.3, 0.11, 0.08], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c286467b8cc106_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.4, 0.13, 0.44], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be2e509f8e9bb081_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.35, 0.63, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca29bbab70b45150_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.24, 0.59, 0.12, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfaa15f085f75efb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.21, 0.67, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40d9a53d498bf186_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.56, 0.02, 0.03, 0.39], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6340560d22234e9a_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_924e149ba2fd20d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9944f84e1c8569c9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.23, 0.48, 0.25, 0.01], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c31fe80e71534ab2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.33, 0.48, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4fcd1f6c69d31e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_012b138e072c65fc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.05, 0.79, 0.12], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1b2f97f93a4a614_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0654d5807df781d3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.61, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3368ead56aed3e7a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.05, 0.44, 0.44], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68f8c01fd4740486_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.09, 0.59, 0.18], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee564c3af3d6e723_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5032a3c2370bb2d8_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0dc56ca0a4e15a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61b977e3c75d95f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.07, 0.26, 0.53, 0.12, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ee3f41ca2a6305_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54f75a3f9c74fd44_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14b6d52777778be2_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b8bb9fb3b8a5a3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72fe386981e43744_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.08, 0.16, 0.51, 0.19, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9cf9e4b41b6ce66_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caa252b01f5074e4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.54, 0.04, 0.38], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fe822e47a9b6628_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.33, 0.21, 0.32, 0.04, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8380e3eca8d53236_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b04a19e77cb02ef_n", "kind": "noul", "options": ["false", "true"], "target": [0.23, 0.77], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93c30bef3767a53c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.43, 0.44, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65d1dc0ca3bc747c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.19, 0.14], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67fff2a3947190dd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.56, 0.4, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd77b18ae74aa459_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.15, 0.38, 0.32, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_379442308dfedbd7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.9, 0.02, 0.04, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5ea568950478a00_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff3613b6a58ada98_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9117a2db159b7828_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.51, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56de72d58106f290_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.21, 0.07], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c53f00553497363_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.3, 0.64, 0.05], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0e67cbfd369cf9f_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66ac037008bc3439_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9ea93bbd2e2e251_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.13, 0.08], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ed61424ba650d3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3, 0.1, 0.58, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_341882cab392440a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.56, 0.26, 0.01], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e23ccf0ca89171d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.71, 0.05, 0.1, 0.14], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbf218dbc273ad2a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.47, 0.4, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc97c5ea9f2e2409_n", "kind": "noul", "options": ["false", "true"], "target": [0.1, 0.9], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6bca7419a25f7c2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0404, 0.8182, 0.1212, 0.0202], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3c6c082b6b09a7e_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d086f0b9364be664_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d933338793dcf62f_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f759b31ca5c1dec_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abf03e6312779830_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.73, 0.21, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b0595540c20cea2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.18, 0.47, 0.16, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32e1d0bd2612c231_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a8ca0ee276120ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6166e74e86e271d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.17, 0.78, 0.05], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9108570beedd5a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.43, 0.47, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7939721339970e4c_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8a41bceddc18878_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c97524cf31e6d41_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.39, 0.51, 0.07], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_791c3e52311209ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.35, 0.51, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ac92e0b890b6ccc_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b025377bb7ed6ede_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98c01fa5c0ae5ba5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.19, 0.17, 0.37, 0.13, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0511140a3ffd75ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_790ac98a21914a6c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.33, 0.66, 0.0], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbccb3ac89d43c7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bafa7e660d147f48_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.71, 0.07, 0.1, 0.12], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1c62223a188078b_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97494f677cbd71b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a71192f72a0dc26_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98a34293cf9d9ed8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.28, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c34d33694f626a94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.07, 0.3, 0.57, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d6f18318826c74c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.21, 0.32, 0.06, 0.41], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6db63a9b4174d3dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.32, 0.61, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ce01c961ebdf743_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c14fa2e5358e4a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b11eb13b5917330_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f102aa5eaa56132b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.36, 0.54, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97eb4ec4af748bfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.46, 0.46, 0.01], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a16603bbadfb5b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.12, 0.5, 0.28, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9782ff5b193e3b74_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.85, 0.14, 0.0], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f48d5bc7d205c154_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2452ff2d94709d86_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1772c9ab9fae2a96_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_609ff41c5b4fe96e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.18, 0.17, 0.46, 0.11, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86ed3d98659418b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.03, 0.31, 0.56, 0.03], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de4d647b19942576_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f80d66a3be05d890_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.41, 0.01], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_406dc854d781f9d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbfb49fc3e225d22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.44, 0.46, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e3325b1fc0d35e8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.45, 0.41, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c46f1b6d2bf47a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b15d0993854f063_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.19, 0.41, 0.35, 0.03], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7045e0903742867f_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41a5483da99db0fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5860f25b7abcede6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.68, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74a03ad9873cae51_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14383c3bf6dfbcac_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2c721106cc5cadf_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e7bbd69c2e4aeea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.17, 0.37, 0.27, 0.05], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_448375f384e5f727_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.16, 0.21, 0.37, 0.14, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f76fc6c53e5831a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.11, 0.33, 0.45, 0.06], "state": "An automation plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9dbba3b96ecb1d2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.25, 0.61, 0.06], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f6cb6c042d1a44c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.73, 0.22, 0.05], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f411c0500168614_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7db077d81572e34e_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9200d8d7b4fd5352_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_545350cc23dba41f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.17, 0.22, 0.43, 0.1, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58234b0cc21bb8eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80b2a5bb76171bd3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.02, 0.92], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b58857e1e41d99c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.21, 0.63, 0.08], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e69099f955dc05b0_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_983264d1eb601741_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0efb6ce86845dab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.16, 0.17, 0.38, 0.14, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_520a7417496df3f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50fc9139606ff7ce_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cec4633d1fc95ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.18, 0.62, 0.11], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e8764d013d6d050_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.8, 0.15, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31502f6df0447055_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.43, 0.53, 0.01], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e330674fcc5ed136_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3939, 0.5657, 0.0404], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13b58a8f3368a0b8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.2, 0.26], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_934430ac92e19fa2_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6377eb523424c5e8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.6, 0.25, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ce3b9d955a7440_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.26, 0.69, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3da706bde7c5d4f1_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4acfb628964955b0_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b1743cf5490abfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.33, 0.12, 0.28, 0.09, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54c228115ecee8b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.54, 0.3, 0.15], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_278e395e631fd4a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_709d241c6cc26a01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.21, 0.5, 0.21, 0.01], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76504429f57f0da7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.91, 0.07, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecf5f18490cd96eb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0505, 0.5455, 0.202, 0.202], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61261446916ba862_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3377480ece924f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4428f4cb876dda16_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.55, 0.08], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17a647b6bf2687b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd7dead90f04356a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.22, 0.42, 0.09, 0.27], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_028039e1e25fcd15_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.07, 0.15, 0.44, 0.25, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff3359e314da8553_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.11, 0.42, 0.36, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdbeba76b5f2de50_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.98, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b77cc6e07a66527_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.83, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_944f6a27f6834fc7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.05, 0.14, 0.49, 0.24, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f40eaca60df888e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31f3f664518c3633_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a39c2c134fbbdea_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd8c8849e04681ff_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.76, 0.17, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2307369a90ee683_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0202, 0.202, 0.5455, 0.2222, 0.0101], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db72b6b29fc32cae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.04, 0.73, 0.22], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4466c7a3fea88fbf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.17, 0.34, 0.43, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3f08e2e0cc40daf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.65, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c4f8af6914355d83_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.38, 0.37, 0.2], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc533a7576106b3c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.66, 0.09], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fad105038a580ed_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.29, 0.14, 0.26, 0.31], "state": "An agent plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce18f8e67c266d79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.37, 0.1], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b679fa9a6ca90b72_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.101, 0.5657, 0.2727, 0.0606], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d7148e693710227_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.25, 0.21, 0.37, 0.07, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dea890377f33e4ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.22, 0.67, 0.06], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f4df05aa49159b3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.19, 0.51, 0.19, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0e23a4df920ebde_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e333d2f212cf3936_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24573e50cc8e35c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09cbc322465869dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d2b96c08abc7ec2_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a10129fefd69e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef939d6075dde301_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.3, 0.4, 0.12, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5576a093ebcc7b7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.53, 0.04, 0.36], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4064e57265f37f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.4, 0.44, 0.09], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dddcb0dd5d8ad5f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.01, 0.01, 0.52], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_548c00fcec078b03_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.5, 0.05, 0.36, 0.09], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_443d27d62a8efe06_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.19, 0.48, 0.15, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d9f1a7e976e1c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3c2ea197d6dd608_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.43, 0.42, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48eec294d5fed903_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.43, 0.39, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eee361e9a9332d0b_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e60b934ebfe6cab8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.28, 0.03, 0.68], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d0432a7e80c144c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.47, 0.04, 0.45], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2f2ccec9f4eb4f3_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96c68a11d667d777_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.57, 0.05], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df128e195120190c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84d635b2065ec8a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_090ed9fa92f06c50_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.22, 0.68, 0.05], "state": "An agent plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2a205b636942a70_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.5, 0.27, 0.21], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0b1ee898b4a1379_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.22, 0.54, 0.19, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd62be36cfb554f8_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5a710bbf06ff00_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.16, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66d77e5905e590bc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.23, 0.16, 0.56], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6459d294028ee5c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.15, 0.54, 0.19, 0.01], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45731559be6b926d_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e28f9bd6ddadb14_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.73, 0.24, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_924ce83fe8655c2a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.04, 0.82, 0.11], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23f0d1c8afa48565_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.11, 0.81, 0.04], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c189c8ab38de005_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.96, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc628b56a69e612e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.63, 0.29], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d4fa72aa5dc730c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.59, 0.36, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_939206d6ffe7eb9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.17, 0.61, 0.14, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_298eb9f451829a86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.24, 0.19, 0.42, 0.07, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45bc355200ca0d98_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.13, 0.53, 0.29, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d84cbec2fcb8285f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.07, 0.54, 0.34, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c4c29b0ba9cd2a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.39, 0.48, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f728f2b26b7fc214_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a61201f80dfb15_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.33, 0.0, 0.02, 0.65], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c51bbd0cdf3ca0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.12, 0.39, 0.36, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bcd316b7e63330a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.76, 0.05, 0.16], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70f55cf635c50299_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbc9344f5decd36f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.31, 0.14], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f08b80af0d4a7c7e_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c7ec41836541945_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.59, 0.11, 0.28, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5f9e3f74bb45b17_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b820ff4de9246dd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.52, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc613c6d0aaf66e4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.42, 0.12], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b11eb13b5917330_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.75, 0.21, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c43b254773c18918_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.54, 0.04], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e9f440ddadb1bb1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.56, 0.34, 0.01], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b06a2c0e4f492a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82e957cdeee1897a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.78, 0.13, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_546701b9a7d0a50c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.64, 0.21, 0.11, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f08d8ac1dc28333_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b7e7d605268071b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.37, 0.53, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf0dab15821bb95e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.2, 0.52, 0.2, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ea69d1f9f3ec006_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08bd68d7df82aed0_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b7c88bbeaa866f7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.05, 0.93], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37c7572491a56f7f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.36, 0.62, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c525a0debd8ed76_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.13, 0.65, 0.12], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd29f42ef59295ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.98, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11a283b71f20ba17_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.53, 0.27, 0.03, 0.17], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2a62b47d5879b03_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.22, 0.46, 0.14, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_767629a49c66526a_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420e4be214e5027f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.43, 0.5, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab3cb8e117d274a5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.4, 0.23, 0.33], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36111af57e0d079a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.46, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efb1ba9965dd9a29_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.32, 0.39, 0.16, 0.13], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18627bd0b5793970_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cae5e8e891d19bda_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.5, 0.38, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_880905eebc7f9186_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.22, 0.56, 0.19, 0.0], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99bfb3bc3cfb3346_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.07, 0.82, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b955160e3396cbdd_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ead24e3fa635dd7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.55, 0.3, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea49a7fc9f04c649_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48b0e98a4a66b49c_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65da0eb8fbba155d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.72, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d2374f68a174632_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.14, 0.47, 0.31, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc193c3d3bd5b5af_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.31, 0.59, 0.09], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0be14e0f640b501a_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a75ef54655d246cc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.36, 0.02], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4953b431025795ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7d30904674b2486_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.3, 0.35], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4099269cc5fa946_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.07, 0.17, 0.45, 0.19, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d33b25c1d7c9dca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.78, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cee5559055499396_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.65, 0.24, 0.1], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1b557da94ab68d2_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b124407b91a3a718_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab73448bd1f6f96a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.26, 0.61, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf55c73888d068ba_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.68, 0.18, 0.05], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3749e6f4bb7c9778_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.19, 0.61, 0.13, 0.0], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13d0b4d32cc778a2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.25, 0.67, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f117c0f0bf56573a_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e038ff0a6931779_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.76, 0.19, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1753f4d3cd5ae888_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.02, 0.26, 0.67], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6107a1159d65aba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.13, 0.66, 0.11], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a213d978aa6d3a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.44, 0.17, 0.34, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95bbd461edd2200b_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664d0708dabb0a71_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.0, 0.98], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47ea5da8dfab6b90_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b610cb03c2449cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82711e257e51cccf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.11, 0.36, 0.08], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f249c7414b621e_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a410b4d9290ae21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.19, 0.19, 0.17, 0.34, 0.09, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7bb9036c6ffd8030_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.48, 0.39, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87b139db602915f8_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02cea0704ecedd4b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.16, 0.05, 0.01, 0.78], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90d6f7529640043c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.12, 0.18, 0.46, 0.1, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f3c1d7e353fdad7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.03, 0.02, 0.19], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d18938c7de5feaca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.08, 0.71, 0.19], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f15d697611568c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.28, 0.18, 0.37, 0.1, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a5a0819e3530752_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.2, 0.76, 0.03], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2752683b0ce259f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.19, 0.17], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_900b95f30c0ca7e8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.68, 0.26, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6c9d1519ddf3c8d_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d95067d6d81c028_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.34, 0.6, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d274b02ad28389b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.26, 0.24, 0.35, 0.05, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_739bf8e6e5842277_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ddcbd1e24b59753a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.53, 0.12], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0d0c4160fa6ef4a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.77, 0.18, 0.05], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0874c321dd891bd6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e87150e41aac1185_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.14, 0.43, 0.23, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d6e350c427ae30a_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f868d122bbc08b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.19, 0.71, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_00d5e20ba0d2e893_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.12, 0.39, 0.43, 0.02], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b04691f0b252d89_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.37, 0.28, 0.19, 0.16], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_718a22a42ac21bcb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.26, 0.51, 0.16, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d54013cfe6c280c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37a423670bdccac0_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95c1a1cb91750dfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af98a9d512807255_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cce7d2fca842f3cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.36, 0.5, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcc6dd27f1235653_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.09, 0.47, 0.39, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0650ef095ae2dd24_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.7, 0.15, 0.12], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fede1f7541a5779f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.35, 0.12, 0.48], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66eeed9d5b1c6fec_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b9290268ccf3f58_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.05, 0.85, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_776b5d10991808a4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.36, 0.05, 0.44], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd7abc4f5bdf73af_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.12, 0.38, 0.38, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21074d093bae630_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.14, 0.35, 0.42, 0.03], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ace44dc492a2e493_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.13, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48a77b7e4ae53d10_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.32, 0.1, 0.39], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_361e46db782ef7b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f16d4bac44712e5a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.31, 0.62, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_722ebc0a44b93a23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.59, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5764e02d4cdce2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_063ff4acd09b4cfb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.39, 0.52, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d70ad19523bd65e_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6cb2ab5bd90fc36e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.67, 0.18, 0.13], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e9be74f53c3e129_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.17, 0.51, 0.2, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e39dd521aaba9bd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.41, 0.18], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b9cbfbb02b29a85_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c757b0e72c2fa035_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.48, 0.44, 0.07], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ea8cfebfdc6919_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.24, 0.02, 0.73], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0c1b9e43a9a23b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.71, 0.25, 0.03], "state": "An agent plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6ff3adf6afd9932_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c708c9f85e4a782_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.36, 0.52, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19b9f3224fc3992f_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d466ad6788286eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.02, 0.12, 0.58, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_090a6a372ae4df15_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_116b28e301cdd6d2_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59261b9bffaab22d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.53, 0.12, 0.32], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_282abe460b40b1c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.16, 0.71, 0.07], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b42523d1ff91c900_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.39, 0.39, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_854019f6b3e8f4f8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.52, 0.41, 0.01], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caef6e13019866f6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.75, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0002c468edcbd614_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.25, 0.32, 0.4], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1c3e2fb1d4a683a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.45, 0.52, 0.01], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e1a80b1ff390637_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.04, 0.94], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9bffd8b4b4c318d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.48, 0.19], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e364f2d6c3b685a_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e7857b00a4b865a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.1, 0.46, 0.33, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_752aa953c24b7b17_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.22, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9cb6eda676554d9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.38, 0.3, 0.23, 0.09], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10bf0fe91ac4b3af_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.68, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01b61491909363ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a72005ffc92732e2_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f5efe4c49dacc0a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.2, 0.52, 0.17, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8cc9324fc7e00a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d79bdaf2c936480_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5455, 0.404, 0.0505], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03912a56a22b3fc6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.45, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53aa6f13658fd53c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.32, 0.17, 0.32, 0.09, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457efabe8101eaa2_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dc5858018cb6157_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc8e3a7d7552371e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.33, 0.62, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0af2813c972afb3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.06, 0.74, 0.19], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef0f862e418e0fb0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.65, 0.32, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7974d1100d8f8663_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.45, 0.4, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_188889171adfd918_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.14, 0.81, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9de3c41efd4c166c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.27, 0.62, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d55a0e9ea5d7c4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f1de1e310fc29a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4afa5095241da70_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5dc9c1a5d0de92e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2828, 0.101, 0.1414, 0.3838, 0.0808, 0.0101], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a39aa2ed422e7089_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.66, 0.16, 0.15], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad47ee9153c37244_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.02, 0.81, 0.13], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26165dd77cb85ae4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.25, 0.13, 0.33, 0.12, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c51bbd0cdf3ca0_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c772aafca72c873_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.04, 0.5, 0.44, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66d2495125e6d86c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.12, 0.45, 0.33, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c36160a175048ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d44f30c48ba2f2fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.4, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d07f61f4f041dc1_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9738231a2e5e108e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.29, 0.66, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17936cebe37839cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.12, 0.7, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c2e042c1af35f52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.13, 0.48, 0.29, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_088451fdb61194d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_708b0c3ae85e9cbc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.05, 0.2, 0.5, 0.18, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_591c92803d1f2387_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8, 0.13, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9039aef5521f6beb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.5, 0.37, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d886a5259de93355_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.78, 0.02], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8319342981a9a54_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.16, 0.69, 0.1], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b960304045286bb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.01, 0.07, 0.64, 0.21], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc8c468571f915b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.13, 0.75, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1061dfbc751d1fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d58df6bdbc8f988b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.37, 0.58, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97eb4ec4af748bfa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.36, 0.05], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_116aa6f2faac581e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.48, 0.49, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca182b778ec05216_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6102ac275a637c01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.37, 0.2, 0.28, 0.03, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8b98792c440eaa8_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8292c1f469b0ad1_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae1c2148cedc0acf_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb76cdd8e743e2d8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.49, 0.43], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae50a32ece50fd7c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.5, 0.08], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_569000e1f3b19cf3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.34, 0.14, 0.32, 0.07, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0537f4fa04f987_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.1, 0.35, 0.44, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8d76212c20f55ff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.48, 0.49, 0.01], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2ed0f23ba28ac64_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.62, 0.31, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9ea93bbd2e2e251_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.05, 0.28, 0.58, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf5d267550558a3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.67, 0.29, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b9b89e991e62f4c_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0378ea7dc1e4f7b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.66, 0.03, 0.24, 0.07], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fa0c5a2f0520acc_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1e8479bba1fccac_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9449ed737dae8b74_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.5, 0.08, 0.02, 0.4], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fffb1a98c8f69ac_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.81, 0.15, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_954ea0f34ed94eda_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c5df943cdaabba5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.72, 0.11, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c20e8f9be4102b94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.31, 0.64, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4280603538002c00_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caefc75f9f492970_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.43, 0.05, 0.41], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_845a76dd016805e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.22, 0.78], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ab5847543a2f6b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.24, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_076763544391681d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.35, 0.13, 0.33, 0.08, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28e7c55063b9a926_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0da13b4795671f60_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.78, 0.21, 0.01], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3993accceffdc9e1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.67, 0.31, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25f2baf8938fa3f9_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c9987e9a8d5c892_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f021521c533974d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.08, 0.87, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a14d0a75b5e6823_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.75, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e91255d77650c15_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.19, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a61201f80dfb15_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec44f25c8cf60d93_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.1, 0.79, 0.11], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40177b253a9b801a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.01, 0.97], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a543b2067975acc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.81, 0.05, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b10fb9fddc3dc254_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.06, 0.67, 0.22], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae350bfe85dd0516_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.0, 0.99], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36378cfdb955dedb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.27, 0.52, 0.09], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e67e0c5a2fc3bf6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.47, 0.22], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44cf965567366988_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7237e44ac0037f62_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1414, 0.2828, 0.5657, 0.0101], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757e864169e8f701_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfd182d9b88e6ab0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53948a31aee00e23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.43, 0.36, 0.04], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0a0489ce3686a4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.04, 0.23, 0.56, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53b6608d177dc514_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.39, 0.36, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d0211e7752a7fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.16, 0.82, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7cb3d9f6bb1930e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0404, 0.0202, 0.2626, 0.6768], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a117425cb1759aa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.75, 0.09, 0.15, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61e4e06bda6333c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.44, 0.36, 0.18], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3749e6f4bb7c9778_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53fadb4a6aa3162d_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f08d8ac1dc28333_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.19, 0.39, 0.24, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaadd861609bbd1b_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0a2c38e41e17699_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.45, 0.5, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25cadd571d3d41f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1fac18724759030_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.27, 0.62, 0.06], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe6f42ea1e457817_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.54, 0.37, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5011ecb83f3444_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933e7b4534733774_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.42, 0.46, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d58a7fdc31dca408_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.22, 0.41, 0.32, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7894accfe1731fab_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f5fc61d42cc823e_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12579b4ec91b9110_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.42, 0.43, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_699fa236d4906e01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.2, 0.52, 0.26, 0.01], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd0441e3b09077c2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.55, 0.27, 0.05, 0.13], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8fd7a5c8194dbfb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.13, 0.55, 0.22, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9368c35590e1c89a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.31, 0.15, 0.29, 0.08, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21de5fbb244ee77_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fbfb624f7c1325c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.32, 0.55, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0bf9390fad569749_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.1, 0.41, 0.4, 0.06, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b8bb9fb3b8a5a3e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.74, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6294a7b212ba4e7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.14, 0.46, 0.32, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_982a3f76b092803a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.47, 0.2, 0.3], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eee63d3463464fe7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.32, 0.45, 0.12, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_745c0d9918654486_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.21, 0.14], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82711e257e51cccf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.17, 0.55, 0.19, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84bd54b888ba198f_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a5a9421a2ad5d9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df128e195120190c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.34, 0.64, 0.01], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13a218bc8aa5a293_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.74, 0.07], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e80e032ed2ff03fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a63899ffbc214fb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.67, 0.14, 0.18], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f97723a78b20cf6e_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fec4778e9660d0a_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7951dd339505fcd8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.65, 0.28, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_726ca9d0b4cddf95_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.23, 0.1, 0.62], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eb20552768160c34_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db44fb6238b20775_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b1b0fa116a38301_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.07, 0.9], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c443dee8ce24b032_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cf61a694b151335_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.53, 0.13], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e0324ad81daffd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.42, 0.3, 0.23, 0.05], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04badb9a4eb770bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91730b7793ec6d78_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5de9d420965a43e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e5740861647e274_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5847daf89d45149_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e801a98765424a7b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.32, 0.14, 0.26, 0.06, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e44f07af9f2bc8c9_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73605d74984c657b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.38, 0.61, 0.0], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76ee4c8c8cd008b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.05, 0.25, 0.52, 0.09], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32f62341fa282fb3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.03, 0.09, 0.45, 0.29, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c7fc57e9d354d06_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.5, 0.34, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6793546761dcffd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.87, 0.1, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecd7cc7c64db8def_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c9e90d4ff10c448_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f028c40ab18d8ad5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.68, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e31b3fd6227e051_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.63, 0.24, 0.11], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35f2fcf7a2deb8e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.22, 0.76], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0be9b0ea271c04ea_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5d004d79878e53_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e72dc475ce846af_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.22, 0.67, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6adc61752a234abb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.17, 0.39, 0.3, 0.03], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2864e21eea12f295_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0, 0.0505, 0.6768, 0.2626], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ade2e6560f3af5_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87bc98fedd0d714a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.25, 0.69, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4afa5095241da70_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.39, 0.48, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d247073ff1bca4e0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.43, 0.06], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a60d7f77eaab32bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.36, 0.14, 0.33, 0.08, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5d2c673372e2590_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.32, 0.58, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42ae43bed99bcbb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de405fdf4136611f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.09, 0.23, 0.46, 0.13], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86290f0a4f672e36_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a9bf3f506759baf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.64, 0.01, 0.25], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a73a02d5425d9a3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.4, 0.58, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67ec0a7b512f3c9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.16, 0.53, 0.27, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_663f119ca52fe746_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b6bb19e6ba706f4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.12, 0.51, 0.26, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51a1a176e1935cbc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.35, 0.14, 0.48], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0dca1c074659eed_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32e5c4d9325d1ed3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.12, 0.19, 0.43, 0.14, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f36243997e53b9d4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.69, 0.15, 0.14], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5f9e3f74bb45b17_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.24, 0.05, 0.7], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a25b3d782f0dd3b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.27, 0.16], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cc89e34840ee539_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44719e17d468f162_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_baeea12bf2d1c921_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.1, 0.64, 0.21], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c0c0713bc560bea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.13, 0.25, 0.43, 0.14, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fadc09bc1f35ab8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.86, 0.1, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7751c6f0cee1c777_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.19, 0.04, 0.32, 0.45], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c0397515dae2b5b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.01, 0.19, 0.72, 0.06], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11fd7eacb93f69c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ebf2aaef76c9104_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.33, 0.0], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_035f33bf819b2408_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acd4c49864576e28_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.69, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53a4bd46613efe83_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_451ab23fadaa6962_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.41, 0.49, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac1b5f3ac7841ec1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.46, 0.08], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f751ccd9e906a623_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.12, 0.68, 0.13], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a42d0229c91bd231_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.14, 0.07, 0.67], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_848e67ded2abc84e_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_807841ad0cc18283_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.5, 0.05], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c53f00553497363_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.05, 0.13, 0.35, 0.34, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d1dd92841725d2e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.63, 0.17], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec6a659e9194cae3_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abf6a0508dde06e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.13, 0.46, 0.34, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfc11fb6b7047bad_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0da13b4795671f60_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.2, 0.44, 0.3, 0.02], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d224c3c24878ac7d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.5, 0.36, 0.02], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f9cef3e41a6bb6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.2, 0.73, 0.06], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0537f4fa04f987_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0101, 0.5354, 0.2828, 0.1717], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_feec5a816ad8c4c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_facd826ee4d3b772_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.31, 0.07, 0.34, 0.09, 0.06], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99bfb3bc3cfb3346_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.21, 0.48, 0.27, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_effb9fe937e0e689_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.19, 0.15, 0.38, 0.14, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25e3f2c6038e1149_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.35, 0.0], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b414af99f6843843_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e12d29a80ecc6c57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.03, 0.19, 0.62, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a491e6482550bbef_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2605f16e337f331_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.04, 0.84, 0.09], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f83134c2ee8dab6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.07, 0.16, 0.55, 0.19, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1ed3b6f9d499ab9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.36, 0.41, 0.15], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_330df55032511b93_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dced71498c2960bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.27, 0.6, 0.04], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ee50dff8b05899a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.33, 0.01, 0.65], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4da32cdbca57989_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.16, 0.71, 0.12], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39d3b654c286f635_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.03, 0.65, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d55c6dd59b69ddaa_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c72fccb90787fa2_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b98afd8c74471948_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.39, 0.14, 0.29, 0.08, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e60c406d72906ef8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.42, 0.05, 0.34], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_945a4ee978383c65_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.58, 0.03], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b471ea2014e4380e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.41, 0.29, 0.03, 0.27], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fcecef046e3518b_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2094f915535298e0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.41, 0.4, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d00825e30076c7a7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.38, 0.37, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed366cb060396997_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.38, 0.12, 0.45], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a5e762fa5c21ad9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.08, 0.41, 0.45, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca3605b67eb614aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d2a5927935b342_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e576acfd138d738_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_516eb18cb3238cae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.26, 0.54, 0.09, 0.01], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f3d1e3f81fdb7ea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.08, 0.16, 0.42, 0.25, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1bcb4199b8cd6f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.19, 0.68, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c96233c729a9e5e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.22, 0.72, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bf784fbd3d66fa0_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_baeea12bf2d1c921_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d92ff1bd865ac6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.66, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fb1a52cde089d99_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.47, 0.48, 0.01], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24cf09a2c39c5017_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.5, 0.25], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ce3022e17dfde6e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.23, 0.66, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b9b89e991e62f4c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.42, 0.09, 0.3, 0.06, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0e67cbfd369cf9f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.09, 0.47, 0.38, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_680974644b4329c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82936cbe38fda291_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b98afd8c74471948_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.39, 0.35, 0.15, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f550d0c687a97a79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.08, 0.13], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6988554bbd5d49c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_886a66b1b1ffa27e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.38, 0.41], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d117fb9c4159689_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82936cbe38fda291_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.12, 0.41, 0.4, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29968b71c569ff6c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.25, 0.11, 0.04, 0.6], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af6878504b38a2de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.12, 0.78, 0.1], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_731695004796ed3b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.06, 0.36, 0.49, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1315fb4846c685f2_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be2e509f8e9bb081_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.25, 0.68, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17d3501a65a8d7a2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.36, 0.51, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c443dee8ce24b032_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.21, 0.67, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e2b2dac639dd269_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.56, 0.15, 0.04, 0.25], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3929c4c582d3edc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.51, 0.04, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f4073e8d1b0d24b_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a219998908c51706_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.27, 0.09], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b3a547817eb4e73_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6246b31361574e6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.2, 0.1, 0.61], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_532fc42cf951190b_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c58da4152a5268ec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.53, 0.23, 0.18], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c54e23bc2145736_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.5, 0.4, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cf925dc454c1e30_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.51, 0.06, 0.42], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bb99492f636738d_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8902342bf666ccc1_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c79b84590c24c1d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.35, 0.3, 0.33], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e2bea7dd5b49f11_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e7d43768b2cbf64_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.03, 0.39, 0.51, 0.02], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6a684036f1696c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.29, 0.61, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_247bfde25b914dcc_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fb1df7329d57578_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb0557ef1b036c47_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.09, 0.16, 0.41, 0.18, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45dc989a35214dd9_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c10747ffb2b406_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.6, 0.33, 0.04], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e1fe486dae0188b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.32, 0.05], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942106c6de437eab_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.05, 0.79, 0.15], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6e41e7cb15f6447_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.23, 0.67, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e63ddfd026f5683_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.06, 0.36, 0.54, 0.04], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efc16bb63a7b65f5_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3f26a340d4defa37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.02, 0.05, 0.29, 0.5, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71f93bc0bd3a0c13_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.17, 0.53, 0.27], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2584bf6b9f2e4354_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d34c97eed7f654d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.27, 0.6, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4538177ef9217703_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.53, 0.16, 0.28], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_233938d0d6441d2a_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d02f8332e63589a1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.72, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16a7daaf8df5a37c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.04, 0.51, 0.44, 0.01], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6610b67c875a471_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.08, 0.23, 0.44, 0.19, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69a37af46c1a8e4b_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd18e184fc1bc1d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.03, 0.38, 0.48, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dff3f4053739b5b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.54, 0.1], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ee3f41ca2a6305_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.04, 0.93, 0.0], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_699c062e3dad48c7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.32, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_885888423eb70e74_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.55, 0.28, 0.16], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c2ab960a710b1b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.68, 0.08], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d07d5b782afb692_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2348123dcbcfe028_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ddc278c3c0f1ddfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.16, 0.71, 0.08], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf5d267550558a3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.41, 0.44, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8f4aa9759289ac0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2323, 0.0707, 0.6768, 0.0202], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b41017e859db61aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.15, 0.51, 0.29, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b63953dd4904fa4_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fb5ac91f58dcfd8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.37, 0.07], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ed4e5dce24d7652_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.66, 0.19, 0.14], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb69727c2c43faec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.69, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c3106032e208ef5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.33, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6923c78ce5427ab_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.15, 0.78, 0.07], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf99b9060c7d84ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_30ede81595d61aac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.05, 0.41, 0.52], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_deb93ee911710c86_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.28, 0.11], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2999fedb4e0322b_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19b8d361d6c58bb5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.15, 0.57, 0.24, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f353d1ffafaa60ec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.06, 0.15, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c2e640271afe601_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.2, 0.13, 0.63], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38d445f75b4e34f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1dbc7b4473888c4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.15, 0.82], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10e7d1c544d628fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.15, 0.6, 0.23, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e70d1f7cda39b5d5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.02, 0.97], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3d45aa139ff2797_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cef36f1b6365d77f_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cf61a694b151335_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.28, 0.58, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdcd58db49a01393_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.38, 0.55, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd1f046209a7933b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.36, 0.21], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3af4f4386837ad98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.51, 0.37, 0.1], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f079978f2f8eadd7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.13, 0.36, 0.38, 0.06], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5237bc2854fddbd4_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ac6f9c5a9b9a09_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.11, 0.68, 0.16], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80e9fc8766359cf8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.31, 0.57, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fcecef046e3518b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.12, 0.72, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a117425cb1759aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.2, 0.5, 0.25, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e303bb01e3307f3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.37, 0.48, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2949f6caf642b23_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94faafcb35e7a3b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.35, 0.47, 0.04], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_127cee9cbff83a52_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_934430ac92e19fa2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.92, 0.07, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa3bfc501dcd439a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.16, 0.71, 0.05], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c4444912d2b283a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.53, 0.03], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd161c26e9cfeafe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.3, 0.58, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52852103f9b467c3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.23, 0.13], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a14d0a75b5e6823_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ce6e4fb5d3fd39b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.02, 0.96], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a822133311d6b6ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70e51c10384431b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c33b0e0b39ed5537_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_961a2f4afce49993_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_189c223d67565875_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.22, 0.72, 0.06], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18f9cdb9edf7a478_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02cea0704ecedd4b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.55, 0.28, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afdd7f59c7aba5c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c2ab960a710b1b_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb0a363886e088a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.11, 0.13, 0.3, 0.25, 0.11], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20e4a2554f969aa6_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df609e0542c72381_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c381678179387e84_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af98a9d512807255_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.06, 0.03, 0.9], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36c1b8f67ec394a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.11, 0.41, 0.34, 0.12, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea6d0c75c9d7f367_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.74, 0.25, 0.0], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a58ec7a7bfcbd56c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.86, 0.04, 0.08], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7628f9945a417fc8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.1, 0.19, 0.4, 0.18, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_489e03428fa7a62a_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fd57dd53243025a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.11, 0.33, 0.42, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e343315796609820_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.1, 0.45, 0.38, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7cb3d9f6bb1930e_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_755967b0d29f3720_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.61, 0.12], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63bde75239d2d16f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.07, 0.18, 0.58, 0.14, 0.01], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a3bee018d02c6d3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.28, 0.11], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_745c0d9918654486_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.06, 0.31, 0.49, 0.07], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb6cf4e29ec47288_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.14, 0.45, 0.28, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_426e86dfb3790819_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.33, 0.13, 0.29, 0.07, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86210a6b8ffc46a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.0, 0.01, 0.12, 0.74, 0.1], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc4537c692ddaff7_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6e8e876991ecf8a_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e3325b1fc0d35e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_751d7465018b3bf5_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68312a8de10982c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc21f7a35e2e802b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.19, 0.37, 0.1, 0.25, 0.07, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d77c8d902337f7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.58, 0.02, 0.38], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f66b13a75b7317df_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c5e45ab8960de9b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.33, 0.34, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_184a39cb430fff95_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39d9f6a46c868969_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.29, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_437894c57f111804_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.07, 0.34, 0.52], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa3bfc501dcd439a_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83544e0b95a7cc8e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.16, 0.24, 0.44, 0.1, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14450ab872b67bbe_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a25336df8d6d89ff_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.58, 0.02], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4663b6ae4b1eedc5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.18, 0.11], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95241ca2e8fd5a82_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.43, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd5087b9da3ae256_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.54, 0.44, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8fd7a5c8194dbfb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.34, 0.63, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9fcb43c43bc1151_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea87355cb7e78d5b_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4410379b943e876_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.31, 0.65, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec044b3fc949240_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.05, 0.12, 0.41, 0.27, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f95babb6c57ab22_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f5df01eb8b81488_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.3, 0.02], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c350bea5065775e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.23, 0.45, 0.28, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_451ab23fadaa6962_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.32, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb35d3e8c0e5c00e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.03, 0.85, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e4c3d3e24670484_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.53, 0.2, 0.15], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_417c9438787afa31_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.11, 0.48, 0.31, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83aee86b399ea3bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.34, 0.53, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccae4b6c9c3c4d3e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.63, 0.35, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ace44dc492a2e493_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.02, 0.07, 0.38, 0.41, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15a2868fb7e6f49d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.02, 0.89, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d10dd1d1cd8d687_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.84, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3841fe515844e2b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.76, 0.18, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8d045508aa2974f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.36, 0.12], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caa252b01f5074e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1bae8ddaf9d153e_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acb10f5da57d6139_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.0, 0.01, 0.58], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d10870c3ba3ec1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec5ea0219f95f73e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.15, 0.53, 0.15, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41fd94c0d9caad07_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_623cf5d202140573_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d291ebf8902aec7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.06, 0.9, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6648b26a85a46e7a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.58, 0.25, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cd901c762c3bc1b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.59, 0.39, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca65d927dec0d85f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.32, 0.59, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a2e024d83329fff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.02, 0.09, 0.65, 0.15], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_446546c30accb6bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.72, 0.24, 0.03], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a979b919d1d3034_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c61c3afb9d90a2db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.16, 0.14, 0.37, 0.13, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8b9388062d1208b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.25, 0.6, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43e79444e0d58405_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.22, 0.69, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d48370daf7eeb9f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.44, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6102ac275a637c01_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.05, 0.66, 0.15], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b536be7294042d8_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c772aafca72c873_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3cb5d98b1cd02f9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.55, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e160dca304097d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.3, 0.54, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_634cf83132e0a41b_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f93c42cd29312c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.46, 0.17, 0.23, 0.03, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb85437c93ff6894_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.24, 0.58, 0.07], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9da8db47b84fe841_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.09, 0.46, 0.33, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a829c511b6f1d38b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.61, 0.17, 0.17], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8e26ce4e14eb363_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8c596727be74773_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4305e5d1d3c06f01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.19, 0.75, 0.05], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8accae430ea8b54d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.48, 0.18, 0.31], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e55ec09ee2560cf8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.91, 0.06], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dff841a5fca40b79_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_873064f89c553200_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.17, 0.82, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cf61a694b151335_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a510160b5e3d3d88_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.32, 0.6, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c88b03890da55bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05ee230f1e4efada_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.32, 0.55, 0.04], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a90ad2af4fd692e0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.17, 0.2, 0.42, 0.14, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4327b4dda1b52600_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b55f23bf3fd4508a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.2, 0.75, 0.04], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc8f2bb182152175_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc76d15a34d3fcc1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.58, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c434137389b39363_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.53, 0.17, 0.18], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9782ff5b193e3b74_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.06, 0.77, 0.17], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_203005c5e6304c45_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.1, 0.21, 0.37, 0.24, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b946cfc07f7090f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.46, 0.36, 0.17], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_975191f5cc8e59fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.14, 0.52, 0.23, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e943f99a67f71940_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_451ab23fadaa6962_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_671ba4dae058fbf9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.64, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc4a322fee89d6f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41d12b2289026310_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.52, 0.0], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8232747b952b2e8e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.75, 0.19, 0.0], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdd1b49bc922fd76_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56108d3783f302f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.42, 0.46, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f03022ac0ca54301_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.37, 0.44, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb6ff29f04e50920_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63bde75239d2d16f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.67, 0.19, 0.13], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed977694b526a4c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_557fbf4a872ae046_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.1, 0.26, 0.59, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed8447c104615b87_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.18, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f943ed77e03fb98_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66820962a5b2331_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f218152faa87316_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.3, 0.69, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57918d94c7408ef4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.33, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98d1c621689639c8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.05, 0.67, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0204c2d3ca675c2e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.66, 0.11], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32f62341fa282fb3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.31, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f64aad792c95d5d4_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a23f465d0db74b8d_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0279ad5eadd45b1b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.57, 0.26, 0.09, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a976ef9d46db6951_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.14, 0.21], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ee2120b0c3593c8_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bf36c409e784f4b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.74, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e1a80b1ff390637_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bbf739a03afcc0e_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8a27da955a31497_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_119a4781b9e3fa2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d54013cfe6c280c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.44, 0.54, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db7d1d92d3a49ed1_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_136c6e0264e153e4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.51, 0.3, 0.18], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_405ee0a1961c919c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.46, 0.19, 0.16], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50b71c9cabc2f712_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.1, 0.03, 0.68], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0ec8d6fc5bbc9ce_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9e961fb3ed33e3e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.46, 0.29, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84b3674e3acfab5c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.22, 0.66, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5301f774082709b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.73, 0.24, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3765236dc0454c22_n", "kind": "noul", "options": ["false", "true"], "target": [0.11, 0.89], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99190a8116f98039_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.11, 0.25], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_779ac12d9306cafa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.35, 0.46, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7167cb4771b91b94_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.6, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd45373dc6ea685f_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b099dafcd653317a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.56, 0.05], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_deb93ee911710c86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.15, 0.36, 0.43, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3f21c31a92b8684_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.4, 0.45, 0.1, 0.01], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a445f519e025b9b1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.15, 0.41, 0.41, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bde3a7faf13bca5b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.48, 0.46, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f04e4153a78a95a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.47, 0.51, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66eab9e11cf1e6fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06c82c93b8b66304_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.17, 0.14, 0.38, 0.18, 0.04], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc96f4f36fd692b3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.1, 0.13, 0.75], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29e20050642fc96e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.43, 0.25, 0.29], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0770e3bc1fd78b6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.13, 0.87, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce11ec06e01636c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0446bac7b6cdd768_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.55, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_910ceb25f05a8dbb_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f04e4153a78a95a_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_898f63d74b0372ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ceb6527e146a8d98_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22892e44c523bf86_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.21, 0.08, 0.12, 0.59], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_888d3188e4b47d80_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.68, 0.26, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40405ee6417b3a11_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.12, 0.2, 0.44, 0.16, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b56d080ec1e3f63_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed04324ea2f1723a_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e4210e866273a7d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.25, 0.43, 0.2, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cedb517c878f3845_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d43940b6c9501535_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.08, 0.43, 0.35, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d8075e8bc272c29_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.38, 0.45, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cdcd58db49a01393_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ce599ca9f636b7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0101, 0.2727, 0.6768, 0.0404], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b4668765d268f40_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.34, 0.51, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_258b0b8ca5f32121_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.36, 0.32, 0.27], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db06280c24cbb4bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04f8e4c0aa4dbdb7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.29, 0.69, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3bf291fc55286efa_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afdd7f59c7aba5c5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.06, 0.63, 0.26], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43ee37918ac34570_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea24c529541e4073_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.26, 0.18, 0.37, 0.1, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6c6141a5aadb314_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.11, 0.43, 0.24, 0.06], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_366cac5f50cbb577_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_529aad22c700fd6d_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c5b244facb79004_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.29, 0.42, 0.16, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34403738e4fc6d06_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9b19509826dc13d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.06, 0.17, 0.46, 0.2, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f578d463866ceab9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.66, 0.31, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e82e732f93e60dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_181eb395ae4ad552_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0202, 0.6869, 0.2525, 0.0404], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c2a597a92a70973_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.11, 0.4, 0.27, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9ab39462e4131ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d75ff1e736c4a8a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.07, 0.25, 0.2, 0.48], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca9a4252b1c46c8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.56, 0.22, 0.2], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce5f7b177feace42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.08, 0.52, 0.3], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db4f2be55cc8df46_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.55, 0.12], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fb1df7329d57578_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.14, 0.68, 0.12], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6000742ec2c9fb04_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.25, 0.13, 0.02, 0.6], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95702db19e5b02fb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.89, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85a64e89a32e5ad0_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a8784d29d315a0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.66, 0.32, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10b1b35b103dded2_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0386211b077ff799_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46478b26415bf170_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.24, 0.22], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bced317f1b3f6edb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.21, 0.14, 0.06, 0.59], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e750dc810a4bf277_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.1, 0.43, 0.33], "state": "A task decomposition ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_760586154e17d289_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.41, 0.4, 0.17], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e721bdaacd1596_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.42, 0.49, 0.02], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85e35b49daf9b135_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ac721995d72461c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.31, 0.37, 0.19], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_682327df6e54500a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.08, 0.53, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cec4633d1fc95ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.65, 0.1], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc854bc563a88d90_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.2, 0.3, 0.04, 0.46], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3ae8d9042fd2447_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.6, 0.13, 0.1, 0.17], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68c4bb1c1eaa6238_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.59, 0.25, 0.15], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcbb903665e0d8a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74341e85d0177cb8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.27, 0.66, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fddcd871c406f308_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.13, 0.5, 0.29, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d3cef507a335d5b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.23, 0.65, 0.06], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_316db6800b7d00c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.26, 0.12, 0.35, 0.09, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d087f591ab7e31df_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.07, 0.18, 0.7], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0bfbf686ffaf14c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.09, 0.51, 0.33, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54f75a3f9c74fd44_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.45, 0.48, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc1c960b7d07b520_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.1, 0.54, 0.3, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25f2baf8938fa3f9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.44, 0.25, 0.14], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac758c61b81e011a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.4, 0.32, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfd8eaabd41a359f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.56, 0.12, 0.17], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_674fa20a6ef65124_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.25, 0.12, 0.38, 0.1, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4703a48beeef821f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.51, 0.42, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea24c529541e4073_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.4, 0.06, 0.16], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420b24b4451181b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.27, 0.56, 0.05], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1acb8a698b381b62_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.13, 0.17, 0.41, 0.16, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a94bedb65a50312a_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_463fd86d4d16bd42_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80d0920bff95cbba_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d92ff1bd865ac6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f091b0967ce7ff53_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.76, 0.17, 0.05], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18a0aa3894874831_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.56, 0.37, 0.06], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c17bc3726cb08408_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0101, 0.2727, 0.6869, 0.0202], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24f1643ba878b640_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.02, 0.34, 0.6, 0.02], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a75c4ef0ae189577_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.62, 0.03, 0.31, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09ede11f5a8fa5bb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.26, 0.59, 0.06], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c76c29fa97f8dfd_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_451ff08e8fd9b65b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.23, 0.49, 0.22, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69a37af46c1a8e4b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.85, 0.01, 0.05, 0.09], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc859c72bde181a0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.47, 0.36, 0.14, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca65d927dec0d85f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.21, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea65180d8b679f93_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.28, 0.03, 0.56], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0be927482a82c71_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a558efff6c01c159_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fba3adb70e0e150_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.71, 0.28, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d271b18350ba6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d7d984399acf0e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d44bd1077964ed6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.08, 0.18, 0.47, 0.17, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c101ad580a7bc7c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.44, 0.16, 0.25], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a16835eef616d23_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2d8b03086a5a292_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.85, 0.07, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a98431bcea23e4dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61ae77c3e1230d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f948e715066f661_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_239bf638e5ec208a_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37a84049a4fbdb8b_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca92046694a00f9a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.55, 0.24, 0.2], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d22f1ffc61c3652e_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c04838e8b5fb06ef_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.6, 0.16, 0.22], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4954a5186e003a23_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3073d1798202acd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.04, 0.71, 0.24], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa9f468c44113d6f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.43, 0.37, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36b7776629c7f6fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d9716fa916e10e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd6ef8b42071c0fc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.1, 0.33, 0.49, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ba6b51a9d789ae3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.11, 0.68, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d86880153e748db2_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fc76ef3c9e8924a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.41, 0.5, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dedb18810907edf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.15, 0.7, 0.15], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ea8cfebfdc6919_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.16, 0.57, 0.11, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0e0547bf2b4b4d7_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20bea058ac26bca9_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_500b156b9f659e7c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.33, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a8b34d4c27413eb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.47, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bf36c409e784f4b_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_904aea898f633d63_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.24, 0.13, 0.39, 0.09, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb46f9e08879ff17_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.78, 0.08, 0.13], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a0a9c4b94bf5034_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.21, 0.54, 0.16, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d2415f9c381d71d_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14754ce6e4f7ece8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.29, 0.51, 0.13, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90db561c14637f6a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.68, 0.11], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b7118f62b93def1_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8802e3457ff96a3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.79, 0.14, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8fd7a5c8194dbfb_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dec118dfd39d53ad_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.64, 0.08], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4acfb628964955b0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.08, 0.89], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ed52b0691fdd43e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.19, 0.73, 0.06], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84b3674e3acfab5c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.52, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_716cf00be3013d50_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41d12b2289026310_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba5f09285e254a9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0202, 0.404, 0.5455, 0.0303], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_511c24ed4e9a3d22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.5, 0.36, 0.01], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_543113fbaf43e0c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51ab43385f943ace_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e6aa0f03e64550f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.15, 0.78, 0.07], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d91fb78fc504c763_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b946f6faa72af7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.12, 0.4, 0.45, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c73ece668808d341_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.01, 0.95], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8802e3457ff96a3_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5118a20b3575ac4_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_299001dabd376d38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.06, 0.36, 0.5, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d866c48604f56bae_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_891b76fb5739a4e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6a13b869121a77_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c39b54f693cd17d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f27d6433a3162899_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9dda360e0fd8f05_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12f86d3bcd3b66ab_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3434, 0.4646, 0.1414, 0.0505], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffb93e578f0f0313_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.66, 0.11, 0.05, 0.18], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c355a3f503bfe39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.23, 0.44, 0.21, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3f3e0bce9df2ae2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.22, 0.65, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee42e7dc67b9d591_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.0202, 0.9394, 0.0404], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb72a0e06056c95d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.36, 0.44, 0.18], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c3597d810b8a9bd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.81, 0.14, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ce42f8ed2f806c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.21, 0.06], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e7ccd7ed60234a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12092eb20a316936_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.23, 0.64, 0.05], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffd5a1454e5e09b8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.14, 0.35, 0.34, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_552c7d8c66cb94fe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.17, 0.12, 0.1, 0.61], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83e6214fba4ab02c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.46, 0.38, 0.14], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad13746280d27622_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3af4f4386837ad98_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a82bf8711f7925b4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.81, 0.01, 0.04, 0.14], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4096de6d789c6d69_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.42, 0.42, 0.15], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d683c63273bdec79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.22, 0.2, 0.34, 0.1, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e13cb220fc6916b5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.86, 0.13, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9387238920d9c009_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5bf992aea4e38392_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab944c47b3d79d16_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_234a075eae0575ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd4c2d46092907cf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.86, 0.12], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f3fad8ffcd7ef50_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29e20050642fc96e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.42, 0.4, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_871c63640aae46fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.44, 0.46, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_707532d73f7bbbe2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.13, 0.44, 0.29, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2e22c8aa30ac982_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.21, 0.22, 0.52], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_458f2468a2ebf546_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.37, 0.58, 0.02], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a73ac89663d2aae_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.39, 0.11, 0.2], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d44f30c48ba2f2fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d7d862a8042f4dd_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61282fd26edc305b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.17, 0.47, 0.21, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_788197a165507557_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.58, 0.29, 0.12], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cf3f424214aca64_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f0d8dd4b2a39e29_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.08, 0.17, 0.38, 0.21, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8a881063e65d1f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.04, 0.25, 0.66, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91255cdbd80c2f7d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.08, 0.4, 0.45, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_443d27d62a8efe06_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.62, 0.23, 0.14], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a7b77027cd5b1f6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.2, 0.17], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5f478b8cf56baf6_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_179afc8b13e9b3fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.2, 0.8], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b767c931f026433e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.43, 0.53, 0.01], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_745e1edf59cc8b8d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.45, 0.13], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8df58a3e170d4473_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_133d70a53cc0117a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0303, 0.101, 0.5455, 0.2626, 0.0101], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd68627f81e01e73_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.36, 0.59, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f0afdd191626b26_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.15, 0.83, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_510ecf4fed3932ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.21, 0.58, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5add67e3ac8bfeab_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ea3b35b2df49a7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.03, 0.96], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22892e44c523bf86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.3, 0.18, 0.33, 0.1, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69ce1c7f0f294e09_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2a3b43cc3735a19_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.4, 0.49, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faefea2f243d1941_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.1, 0.8, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a49e867ad3c4b42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.09, 0.16, 0.47, 0.14, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e1671396a87004c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.06, 0.56, 0.33], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37eb72145386aaac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.34, 0.51, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74341e85d0177cb8_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e7ccd7ed60234a0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.21, 0.46, 0.18, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d93c13872898063_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.23, 0.65, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7214566c4380449_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.13, 0.69, 0.12], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_622a3b2ec7c5d457_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.76, 0.14, 0.08], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9b0e6754bff5807a_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6fd47ee23167d52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.53, 0.04, 0.41], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9bd2d7ead6a2904_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.06, 0.42, 0.45, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbbbe0432ff72e52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.04, 0.48, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757ec56d5379abce_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.66, 0.31, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df38c381467cbed7_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_250b9a909f2c78c4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.73, 0.16, 0.1], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c17f0d955e8c034e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.48, 0.44, 0.06], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fade8401009d8906_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.0, 0.03, 0.92], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e55ec09ee2560cf8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.05, 0.12, 0.47, 0.28, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe93e33a6d486c37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.16, 0.7, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cfadf07cbf2a7ab_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f27d6433a3162899_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.25, 0.63, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3143aa8812ab15c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b476103258e042c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.42, 0.43, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d555f8cb2726d65b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.34, 0.49, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd7092a1e7f9afc6_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ffd73d6f3643cd7_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cdfad2ec70a5c11_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.89, 0.08, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2123b8a1aed0e64c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.33, 0.51, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acd4c49864576e28_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcf3a0a5f2ce7e04_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.97, 0.01, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3aca263e8d939555_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.4, 0.23, 0.36], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8a76aedf569fbbf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.04, 0.53, 0.42], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d1dd92841725d2e_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3d45aa139ff2797_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.19, 0.76, 0.05], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f797452272707f42_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a059da170099a97_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.29, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e47fce249867845_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.23, 0.65, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36111af57e0d079a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.19, 0.43, 0.33, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f5df01eb8b81488_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fff5f2504c696101_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.11, 0.83, 0.06], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3841fe515844e2b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.15, 0.46, 0.27, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17906a7a7c09badb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.01, 0.96], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbc9344f5decd36f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.15, 0.46, 0.27, 0.02], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cac519fc57cb0e0d_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c0ac10088c5b0e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d03f29a753f8f9ea_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.23, 0.19, 0.55], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7cd5b69df616bef_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76d316adfcf10e8f_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9944f84e1c8569c9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.0, 0.98], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fa80c99d27ff915_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.11, 0.82, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d683c63273bdec79_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a609bc291f7c944_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.59, 0.2, 0.18], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3a7106f06421323_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8fed5a99b63e233_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.84, 0.15, 0.01], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1c6099a22aa3e5d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.06, 0.76, 0.17], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dfcf196449abdf3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.6, 0.09, 0.27], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94571f9e2128dee7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.16, 0.23, 0.56], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d4e650fbd5c00a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.08, 0.39, 0.41, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_021b478b27fbe3ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.69, 0.22, 0.07], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0204c2d3ca675c2e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0202, 0.2828, 0.6667, 0.0303], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f790be79e1c45d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.18, 0.76, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c4fecb78131315d_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1504b013343c317c_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05c83edcd842ad19_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.36, 0.55, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc36e6bbb370c05a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.07, 0.22, 0.55, 0.09], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76d316adfcf10e8f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.33, 0.03, 0.01, 0.63], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0dfb4a26c86e8b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.37, 0.39, 0.07, 0.17], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5100fac467897a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.1, 0.53, 0.35, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c5218b87ea0fb89_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.54, 0.39, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883c2559626f4e83_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.07, 0.48, 0.41], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86a2054841463def_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.2, 0.16, 0.36, 0.15, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7a59cdb0515a9e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.05, 0.25, 0.57, 0.04], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e56fba75e395d19_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.1, 0.36, 0.46, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d92fe5c48a0c3161_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.03, 0.23, 0.65, 0.04], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3ffd7b31c8ae86b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.36, 0.08], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10bf0fe91ac4b3af_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_910ceb25f05a8dbb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.15, 0.09], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5498145fe6a54907_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66ac037008bc3439_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.09, 0.75, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56a9b45a7d9daa5f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.79, 0.18, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc3026ef0f2905cb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.47, 0.23, 0.06, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_342ff9f6d8974b86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.17, 0.57, 0.22, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad47ee9153c37244_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.08, 0.26, 0.43, 0.18, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfec556ccaeaea5e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.38, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_845a76dd016805e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.47, 0.31, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d0432a7e80c144c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.19, 0.19, 0.4, 0.08, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a381ea4de8b66554_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94571f9e2128dee7_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66b4f5086c99afff_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.08, 0.54, 0.35], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54a5968f9c2d26c9_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8eb765f46752ad1_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e2e33110b8a1a2d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.76, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9f30ef8c32842e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.32, 0.08, 0.32, 0.11, 0.05], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8806d4268e4227cb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.26, 0.37, 0.3, 0.02], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a6e74bd345b9aa2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0606, 0.0101, 0.0101, 0.0808, 0.5657, 0.2727], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_182fbf9911ed2edf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.04, 0.38, 0.52, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13d0b4d32cc778a2_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3f26a340d4defa37_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.25, 0.73, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31e1bfdefc81e199_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca98020acc81c7b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.49, 0.2, 0.11, 0.2], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f15ee87d9c20b25_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.4, 0.51, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0fca341cc136a09_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.41, 0.21], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfecae896770d57f_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7a9557c57aaf356_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.16, 0.48, 0.2, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5820ca6ea318f88f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.33, 0.52, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da58e7b489dd4554_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.41, 0.1, 0.3], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b960304045286bb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.78, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9b1a0cabe910127_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a5a0819e3530752_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.3, 0.54, 0.13], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16949c5f70f32545_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2057bedcf356e9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.32, 0.1, 0.37, 0.07, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e404304adb4ba26_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.13, 0.5, 0.34], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2dc2e07037e910c8_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37b9b95bd48e8abe_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e7857b00a4b865a_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dfa46782de59924_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59c41fc94cfd5832_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.82, 0.16, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0313e7dd945e2466_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.21, 0.05, 0.49, 0.25], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c286467b8cc106_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.11, 0.15, 0.44, 0.15, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01cd0da562c8c559_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.5, 0.43, 0.01], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ca067c8d7da4c0b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.22, 0.5, 0.1, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1557e3813706c2bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.07, 0.93, 0.0], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6c4d58182afc6e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.32, 0.39, 0.09, 0.2], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd7abc4f5bdf73af_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c828ca053ec974f_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415da395b7b74348_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.29, 0.46, 0.23], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ba7ad1b503e9233_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.28, 0.67, 0.04], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea1ead63b89e9259_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.79, 0.16, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da2a643214ced61_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.61, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e2b2dac639dd269_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.17, 0.16, 0.36, 0.15, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5958de6facfddd23_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_175128c07b0e1b97_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_172c23b0449e0268_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.54, 0.39, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c1cd571d2707e14_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_07040ed65fd98c14_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.44, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_271d0ebe45be9b70_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.58, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6294a7b212ba4e7f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.37, 0.48, 0.14], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0932dac066292b13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.12, 0.7, 0.12], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13be773c0297e86d_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d555f8cb2726d65b_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1268d0b86551ff3c_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8319342981a9a54_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.22, 0.49, 0.29], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ba7ad1b503e9233_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.33, 0.0, 0.0, 0.67], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ebf2aaef76c9104_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea65180d8b679f93_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.24, 0.19, 0.35, 0.11, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8910c14f60798a2e_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec38a1a0086734c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57a6cfc3de4e31ff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.14, 0.23, 0.39, 0.15, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6555216168a7a868_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.08, 0.03, 0.79], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_767629a49c66526a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.32, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c251f31c548527da_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.35, 0.6, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cb9d90d0bcbdf1a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.19, 0.66, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b9515c4826bf534_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0101, 0.0101, 0.0404, 0.9394], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64c7d221ad3f7656_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.34, 0.59, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0960237b382aa9a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.11, 0.31, 0.41, 0.08], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3f61057825835aa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.67, 0.03, 0.13, 0.17], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_618dbd980e02fab5_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae591bd08c92119a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.58, 0.24, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc2fc02338f02fd6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.21, 0.7, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4bbf739a03afcc0e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.42, 0.22, 0.33], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a49e867ad3c4b42_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6375e93f2f6ae957_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.15, 0.37], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a715b907e4a223ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8e19d2f422c039d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.31, 0.13, 0.36, 0.07, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cd29498fb8de605_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9e1908f31f254e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec1408e7cb758df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.18, 0.53, 0.21, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_579e8bb0bb2860d1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.33, 0.56, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fe822e47a9b6628_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.51, 0.29, 0.08, 0.12], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f853e4d570293112_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f1701b1ba64bab9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.01, 0.96], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be7c71801ea27565_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.24, 0.03, 0.01, 0.72], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_097b94e33ccc4a83_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.41, 0.56, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2bcde3b10834015_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2ed0f23ba28ac64_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.01, 0.18, 0.71, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf0dab15821bb95e_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad062133d0c325ea_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.18, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a850d0a86d295d1d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.3, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0231052b5cedaf4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9932ac390479b8db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.07, 0.32, 0.5, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fddcd871c406f308_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.46, 0.48, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d734daba35e30d51_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.19, 0.04, 0.41, 0.36], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c1e123c10a72a48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.45, 0.04, 0.26, 0.05, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5684d44875c68eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37f8afca792ff0f4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.61, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93f28a2d641d2549_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.19, 0.2, 0.44, 0.08, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f39069fb8fe8e4c_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cef36f1b6365d77f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.08, 0.81, 0.09], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd77b18ae74aa459_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98a34293cf9d9ed8_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b4eecb6fbf12d6e2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.01, 0.97, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31f9f37481d2d97a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.45, 0.21, 0.21], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c0ac10088c5b0e5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.44, 0.48, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9f709e7037ec7d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8662b4fa585b027c_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc43a07e0586044c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db7d1d92d3a49ed1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.16, 0.39, 0.4, 0.03], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bd7ce93aa5572d3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.65, 0.2, 0.12], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b979252c26119420_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.19, 0.71, 0.08], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277bf9874476a3aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.24, 0.19, 0.36, 0.07, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f48d5bc7d205c154_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.14, 0.7, 0.11], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_543113fbaf43e0c6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.32, 0.61, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_435dc03e02aab303_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.11, 0.65, 0.2], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51a1a176e1935cbc_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98a34293cf9d9ed8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.44, 0.48, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51f2fa5f7db0b5ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.21, 0.43, 0.23, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d892f2b26a15a333_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.14, 0.39, 0.44, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14450ab872b67bbe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.15, 0.78, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a7cde69c202d5fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.34, 0.32], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7951dd339505fcd8_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6a43b48e74e2fde_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.83, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_522127d569875724_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.2, 0.27, 0.48, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b3edef5da93d2b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6d55bd8c57867b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0101, 0.0303, 0.404, 0.5354, 0.0202], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_786a51f0cd17b110_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d921cc5a6ba9c4a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82936cbe38fda291_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.39, 0.08], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c11ca2dc14e6cc3d_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b095f0998c5a371_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.25, 0.64, 0.06], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8858b20ce5f77ec7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.91, 0.02, 0.04, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5ecdcf36665ade4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.03, 0.75, 0.18], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6e2936559b2bcd5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.19, 0.43, 0.07, 0.23, 0.05, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfd88908d0ac28ff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.23, 0.63, 0.07], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac11a0e4c3654680_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.08, 0.77, 0.12], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee0afb9e90934c11_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.47, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e23ccf0ca89171d_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d92fe5c48a0c3161_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_527603a833da6ea3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.8, 0.13, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33046db768a0b74e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.34, 0.49, 0.04], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36356e38be922cb7_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e113e6827c63974_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3c2ea197d6dd608_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.43, 0.15], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f35ec43466addbfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3f400661d9b30f72_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.34, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f95fe8772f97f41c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.11, 0.66, 0.17], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56f99c8bb8de7d43_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.61, 0.25, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9e08cb9d3dc50f2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.04, 0.34, 0.16], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd476f30e5788c91_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.17, 0.48, 0.27, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f4c42ddab15b25c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.54, 0.33, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f592711a691fa547_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.06, 0.9], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bfcc172cce29480_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.47, 0.23, 0.29], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4941af120a72dd48_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80025f09c0bc07aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.26, 0.58, 0.06], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a976ef9d46db6951_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.28, 0.52, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5dc9c1a5d0de92e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.69, 0.06, 0.22], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6ff94256d3d1abd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.19, 0.15, 0.36, 0.16, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4d1859cfa174a01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.21, 0.45, 0.16, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b4dac0b8f1fce94_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.91, 0.03, 0.02, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05fe3bf52e6d1e57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0101, 0.1919, 0.6869, 0.1111], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dee3ad473e590677_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e8131ec88b99f99_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94498683df8fe79b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.42, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90b30e4527c34f8e_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e29dd15a942b6af_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9efba9c1dee4c6db_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.44, 0.4, 0.14], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_726fe42b823fa468_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.25, 0.65, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21e6e4c0e6933e23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8, 0.12, 0.08], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67aef4555f5e46bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.38, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd441ac060bc3d0d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.1, 0.32, 0.46, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a558efff6c01c159_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.45, 0.5, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8074deb77843f9db_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.51, 0.38, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d274b02ad28389b_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18428a8347470f38_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.02, 0.95], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2dd1df6151e9a4c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.17, 0.83], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d00b08798b0de41_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.23, 0.73, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fb1a52cde089d99_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.01, 0.12, 0.82], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4466c7a3fea88fbf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.64, 0.33, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31dfc80f7da53384_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.14, 0.47, 0.25, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_639f06566c041fa2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.31, 0.5], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9f709e7037ec7d9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.53, 0.14, 0.31], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d224c3c24878ac7d_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ace53cfc46edd7a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3, 0.01, 0.14, 0.55], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02c9eef1af87fcc3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.69, 0.12, 0.14], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac6285784d4a38c2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.37, 0.12], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b476103258e042c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b113691808a2ec60_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.31, 0.01], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10ad3427799710e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.03, 0.83, 0.14], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fb2c8fc2ff88647_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.39, 0.51, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be6bb4e90c01ef57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.08, 0.1, 0.35, 0.35, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35c4096348653c21_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e4ddf6b03025a86_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75c15ec4f8364337_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.87, 0.12, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b79b625ceeb976c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.31, 0.55, 0.03], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49cce07ff942abd2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.16, 0.76, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be7c71801ea27565_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.22, 0.08, 0.35, 0.17, 0.1], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51cb60461e068ac2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.13, 0.45, 0.35, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2803dea13d55c33_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec5ea0219f95f73e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1, 0.44, 0.11, 0.35], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e55661623d6a62_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.24, 0.18, 0.3, 0.12, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dd511831a507dd5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.17, 0.08, 0.72], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee53640af2d62757_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.08, 0.07, 0.8], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_527603a833da6ea3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.18, 0.55, 0.19, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47aebb6cfa6a43b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2878faf404c1885_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.96, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2e4be0ebf455f30_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e96b6b16b7999f38_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.68, 0.13, 0.1], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcc1c231722a9fd7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.17, 0.17, 0.42, 0.11, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b255d7a2a2b38490_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.24, 0.66, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40c7fcd15d0b98b3_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a4c3fc57e37a4e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d0211e7752a7fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.15, 0.76, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e2fe20b52051f43_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.49, 0.22], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37eb9a0fe1166b88_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.28, 0.5, 0.07], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ec27ee2c7c667d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a25b3d782f0dd3b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.23, 0.44, 0.14, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4663b6ae4b1eedc5_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ae1e9a779088097_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.63, 0.35, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f759b31ca5c1dec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.34, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de146d67ec369d6c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.17, 0.18, 0.47, 0.11, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f79ad6dc79db39d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.18, 0.52, 0.18, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6013d43e08494502_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.12, 0.53, 0.26, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0cee2e5e9c7f165b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.09, 0.38], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a1c98637a87531a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.59, 0.34, 0.01], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_172739dc8dc5d65a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.21, 0.18, 0.34, 0.12, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44d5b4bc290ef385_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.22, 0.18], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd5087b9da3ae256_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75ae17bbae613c90_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.07, 0.21, 0.55], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_006adc6f46720911_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.81, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c1a7e63eb1d94c0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.43, 0.05], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3bbb8f61ba3727a_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56d1c8335e745f07_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.14, 0.67, 0.12], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5945438dde376d91_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.17, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5d8eb7c65766023_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eba09f30847e5e78_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9b1a0cabe910127_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.17, 0.45, 0.3, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_506aacd399e08900_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.03, 0.1, 0.48, 0.28, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d336a11db32d032_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.05, 0.63, 0.25], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd161c26e9cfeafe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.39, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_233471807e869217_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.26, 0.23, 0.37, 0.07, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_511f2ba43686a1e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.31, 0.66, 0.02], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c39b54f693cd17d5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.35, 0.04, 0.23], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3f08e2e0cc40daf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.19, 0.7, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_539f5a46052e3b65_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d94256da9255c18_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0606, 0.0101, 0.0202, 0.1717, 0.6667, 0.0707], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_959b4bb761f9fd3e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.33, 0.22, 0.13], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_03093030590dd0ae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.19, 0.23, 0.37, 0.11, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8c6662a2840fb88_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.1, 0.5, 0.34, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fa80c99d27ff915_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72c3f40d6f20384d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.45, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_203005c5e6304c45_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b08e91cbab6c2b13_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.04, 0.75, 0.2], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c44f4ac71e31ba9d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.08, 0.14, 0.43, 0.19, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c75832a3ac148a6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.33, 0.05, 0.17], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cc89e34840ee539_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.15, 0.2, 0.39, 0.13, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c525a0debd8ed76_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.68, 0.27], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b08100c3159e914_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ec36d785c6015ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.13, 0.53, 0.2, 0.01], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_adce541271349f46_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.42, 0.09, 0.38], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9ef3836e42bf110_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da1a7e66012e69a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcb6a245ed646125_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a2397e9e6ca9b2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.29, 0.56, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab4e4a897a07f62e_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4096de6d789c6d69_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2999fedb4e0322b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.09, 0.3, 0.44, 0.14, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f970fb3f95896efe_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d84cbec2fcb8285f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.47, 0.02, 0.37, 0.14], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcbfc548f1f66d1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcf267c86ee995aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.15, 0.41, 0.35, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b949182670528683_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.08, 0.15, 0.41, 0.25, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e1f7a2109ed89c1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.59, 0.23, 0.17], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a491e6482550bbef_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.8, 0.03, 0.08], "state": "An automation plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f86866d2979e760e_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b04bdcb096a65ae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.38, 0.42, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_78649bbe4e9d56e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebabd95fa4217fc1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.61, 0.35, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c229aed5289b89cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7afdf61d85fad916_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.09, 0.41, 0.28, 0.06], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1441e1a2a739762a_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06ebc2dbe9754df8_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f3fad8ffcd7ef50_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.15, 0.43, 0.28, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51a1a176e1935cbc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.14, 0.41, 0.3, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ee2865018d36731_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b4ce1c46bc3adb2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.63, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b01d579bfb13a303_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.46, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8bc25af0e15e819_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.12, 0.7, 0.12], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533c5b71ebce21b7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.47, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9412a09155e578bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.42, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0334399392b60f51_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.79, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc628b56a69e612e_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc0270a376d49151_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.21, 0.68, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a2a800ba6d09a32_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.06, 0.16, 0.47, 0.2, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba8108e3e0d7b1c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e113e6827c63974_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.1, 0.59, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a97a1b2f37eb1b5_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b18e67555da367b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.03, 0.01, 0.65], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b472344d7024138_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0f6fd65e2411611_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.35, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1105cd2e76c98f48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.09, 0.42, 0.38, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d336a11db32d032_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9108570beedd5a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43696ef77c88edbd_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39c23f5624930a35_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.15, 0.7, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93bc2f37bc45a28d_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37c7572491a56f7f_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f77ea8ed855743db_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.65, 0.2, 0.13], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c40d20c8e407380a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.17, 0.37], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f746f00bb1d89263_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.17, 0.47, 0.17, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2ccc94a7d2d64c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_945a4ee978383c65_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f98acc8fb667ed5f_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b62885408dae16ec_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.36, 0.19, 0.3], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acb3c2921c7a242c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.03, 0.7, 0.19], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7ebac4fc92e459a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.42, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_440a473d3c30c59a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.51, 0.45], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e404304adb4ba26_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40532f286c66359f_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2102be411c2f55b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.81, 0.18, 0.01], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1b557da94ab68d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.41, 0.12, 0.27, 0.08, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15cf2fde77f59b21_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.88, 0.11, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a49c31df3fce37a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.12, 0.84, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcf3a0a5f2ce7e04_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.22, 0.42, 0.04, 0.21, 0.06, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f249c7414b621e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.7, 0.06], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47aebb6cfa6a43b7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.34, 0.06], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42b2e86640c7c4ac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.05, 0.43, 0.46], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90f8845d79f24995_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7beb6c6172b153ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.12, 0.16, 0.45, 0.15, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ae43a6a43f7f434_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.32, 0.0], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e017d0eb649cb055_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_07c6769c5d1dcac7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.84, 0.12, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12579b4ec91b9110_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_677bbc368d055a9a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.17, 0.48, 0.17, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdaed520e703c97b_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_747e047521520a7d_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41b0f7e09579be93_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.14, 0.53, 0.26, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba5f09285e254a9c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.37, 0.01, 0.04, 0.58], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4cfbab243389aef_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.45, 0.24, 0.14, 0.17], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85e35b49daf9b135_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.02, 0.95], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ee15a08a5c44163_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa48d936fa4241f1_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d13601f00d11c09_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.12, 0.38, 0.31, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b50e29c4e24db95_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.14, 0.45, 0.19, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74487a1b946d1919_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.4, 0.23, 0.34], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59a2d242b0a74d68_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3094dd51c925df54_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.19, 0.44, 0.19, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99bfb3bc3cfb3346_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b993ff2a040bfc64_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.33, 0.62, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e109a1030ed525f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.17, 0.43, 0.28, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c21709c7047ec6f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.29, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd791bb12d3ff35b_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a9e977023f1a4e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.07, 0.14, 0.53, 0.23, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3753817c96e6034b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.15, 0.46, 0.22, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0dc35d3ebe9f750_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e2b2dac639dd269_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fad105038a580ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e866fb21121eb12_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.64, 0.04, 0.1, 0.22], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d820d82dd055280_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.12, 0.13, 0.45, 0.17, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57c9825def0e7a2a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.46, 0.33, 0.06], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a9761704d49df88_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cccf80115099d7be_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.58, 0.29, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c51bbd0cdf3ca0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.15, 0.79, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb2f8bf739786db9_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4d1859cfa174a01_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1872980fefc17814_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.22, 0.51, 0.19], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd23192908e658ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd8c8849e04681ff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.07, 0.4, 0.42, 0.1, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72c1790e619df6be_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cad86a6e0f307578_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.05, 0.81, 0.14], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1401d87832234d3b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.47, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f9230f1146f55ef_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.58, 0.22, 0.18], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d14e6d5fbcd6dfa0_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_502d8a6158c2f31a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.02, 0.12, 0.65, 0.14], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ee50dff8b05899a_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e2e33110b8a1a2d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.56, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9bdd34ecc6fc334_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dbccb3ac89d43c7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.11, 0.36, 0.49, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c9b91d17bed563c_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2878faf404c1885_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_559773473da5fd9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.3, 0.54, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afc380fc182a620b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.22, 0.72, 0.06], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e0f550f12a0a006_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5be7b99a039592c_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39dbb8fce4ff9aea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.06, 0.48, 0.4, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d555f8cb2726d65b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.81, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_155492440990c9f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.11, 0.22, 0.39, 0.16, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_421705ea38b712d0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.13, 0.72, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92f4fe645926f762_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50fc9139606ff7ce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.27, 0.57, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69b25b785070925a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.15, 0.55, 0.26, 0.01], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5ecdcf36665ade4_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61261446916ba862_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.08, 0.4, 0.4, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd791bb12d3ff35b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.44, 0.08, 0.19], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0a2b29cf61d3b1e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5657, 0.1616, 0.2727], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f0589a1b10f30e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.18, 0.68, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0d6fe8cfd881e51_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1b9064409908d7d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.73, 0.03], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c007c734adb75dc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.1, 0.59, 0.29], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04badb9a4eb770bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.05, 0.36, 0.5, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_617188c23d5bf8b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.09, 0.38, 0.44, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c101ad580a7bc7c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.17, 0.42, 0.22, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cd901c762c3bc1b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.39, 0.52, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8efa5db44975027_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.79, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d1d48ab06b690cb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.2, 0.77], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_713fdde29303229a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.11, 0.34, 0.5, 0.03], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1890a59accfe573_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.2, 0.74, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4920a534048d41aa_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1bcb4199b8cd6f2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.37, 0.61, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a9a349005728f0c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.25, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b1f4d068952d6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65a29a7094c8f953_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.02, 0.87, 0.0], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e12d29a80ecc6c57_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdc84e300710c66b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.56, 0.29, 0.14], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40c7fcd15d0b98b3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.58, 0.01, 0.36, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7a2244760cf3c3f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.3, 0.43, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56c4064674cedd40_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.39, 0.51, 0.02], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58234b0cc21bb8eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.13, 0.38, 0.39, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae93774209e1d29f_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3684caf19a02cec4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.22, 0.42, 0.3, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_366cac5f50cbb577_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.62, 0.34, 0.01], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415b32b9c52d106f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.22, 0.6, 0.11], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf4c7bf0d295d29e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.24, 0.72, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6988554bbd5d49c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.29, 0.58, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dd511831a507dd5_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933689cf11261639_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.37, 0.48, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5804b1ea89d925da_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9934de50556b441_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.15, 0.43, 0.24, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34403738e4fc6d06_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.49, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ee4f1b4d9c8ca45_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21074d093bae630_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0705f5eff2ca22e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.46, 0.43, 0.02], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5396adf9bbbd700e_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7feba9ca0e8f0abf_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88fee1bbfc86fbbe_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74ec8b81ed2830f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.42, 0.46, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0d0c4160fa6ef4a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.22, 0.61, 0.11, 0.0], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e9aaef111397c54_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.36, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58fb3ead4216207e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.25, 0.73, 0.02], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ca91996a5bcc4e_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0313e7dd945e2466_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.26, 0.38, 0.29, 0.03], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_800c2e193528ce37_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.25, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9bca918a7877c21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.01, 0.04, 0.55, 0.31], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dac15cfc1703e03_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.73, 0.24, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bed8d226119c25b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.1, 0.39, 0.37, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d83fe27cb45f5f55_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dabd819e65a9b59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.22, 0.7, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afc380fc182a620b_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfe52f0bbaa41d4e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.05, 0.6, 0.26], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbe2cb93d6ec0290_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.29, 0.69, 0.01], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dfcf196449abdf3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.14, 0.43, 0.34, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_871c63640aae46fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd0d86ebb49f4451_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.24, 0.17, 0.37, 0.11, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_072d65f96448fcce_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.52, 0.02, 0.38, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25d2386ceb990356_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.2, 0.68, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e79d165c8e87629_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.18, 0.7, 0.06], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c10747ffb2b406_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.32, 0.49, 0.13, 0.01], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_522127d569875724_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.23, 0.22, 0.39], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c63a9035bb816cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.17, 0.46, 0.26, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c278ccf5b704ad23_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6f91126611ee44c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.64, 0.16, 0.19], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933e1d34a769fa51_n", "kind": "noul", "options": ["false", "true"], "target": [0.23, 0.77], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a7b77027cd5b1f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6563add88809eda7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.02, 0.3, 0.58], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bce8cdb655f6892f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.16, 0.55, 0.22, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5485a116a01bed_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.02, 0.51, 0.4], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfd8eaabd41a359f_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f1ef042f62760d6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.02, 0.34, 0.6, 0.02], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7be40e18def55d85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.17, 0.64, 0.1], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_014932068facba82_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a36bb7a7287eeb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.33, 0.14, 0.39], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5f56fab736f4fd6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.13, 0.17, 0.5, 0.12, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6231276c852ae808_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca385d03bbb8b60_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fb2c8fc2ff88647_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15213e83e77648c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff682b592e24c80e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.47, 0.07, 0.4, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dc6657cdee76681_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f329d35e680482d2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.48, 0.39, 0.11], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96b1c32131d4b147_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.21, 0.18, 0.59], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b5132c68f7d83e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.05, 0.32, 0.39], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6fa2d67bfee2335_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.32, 0.03, 0.62], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d9716fa916e10e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0101, 0.8081, 0.1313, 0.0505], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a403932a8c09a3e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3f21c31a92b8684_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27461670616936af_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.12, 0.47, 0.38, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_661bab7f5225c8fa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.12, 0.21, 0.48, 0.13, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c505836d0c959b8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.45, 0.05, 0.3, 0.06, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff30d71d81f121bb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.26, 0.63, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_601339f43d2072fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.41, 0.46, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8ce053bf9ae403f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.63, 0.1], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64eb30296601efdd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.25, 0.11, 0.33, 0.13, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_258b0b8ca5f32121_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.15, 0.41, 0.27, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a96d3994cf8175b1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.4, 0.39, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b095f0998c5a371_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_576083f6f0f02f51_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.05, 0.1, 0.31, 0.39, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_899a1381f91590d6_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ca5a6d239515c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.07, 0.49, 0.32, 0.02], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9338336816543274_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.08, 0.29, 0.52, 0.09, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d498cb39ce5f3e34_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.44, 0.54, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1261402c537bd438_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.15, 0.42, 0.29, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6193b0bc82dcfba8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.11, 0.18, 0.42, 0.15, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edb14bb1e22c861a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.23, 0.75, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a4e159c9fa188fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.2, 0.19, 0.35, 0.13, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_592aec12475ac01e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.04, 0.43, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8a687bbb754f25c_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8fc72382c3f8da4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.47, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da943e8d70859ba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.35, 0.05, 0.22], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420932e343b3f53f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a84e85eb988ff68_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.12, 0.29, 0.45, 0.09, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecc3e9af2b753e35_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_487fec480332eedd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.25, 0.53, 0.17, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b099dafcd653317a_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0f6fd65e2411611_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.36, 0.5, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff1652fd8d7755ed_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.09, 0.79, 0.05], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_483530cfc0e92840_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0303, 0.0909, 0.5657, 0.2626, 0.0101], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da366145260381ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b93c73174df2fcb0_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63baae22fbb14cbb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.26, 0.18, 0.31, 0.12, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2452ff2d94709d86_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.63, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bda9e3e06c10134b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.52, 0.41, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d437029cd105254f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.54, 0.09, 0.28, 0.09], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37b280193138c49c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.21, 0.7, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc656df096979d57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.43, 0.47, 0.02], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7d53b7924953a6f_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d3bd4bcb1ecd027_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44fc15661deebd59_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_990cef6bc329c560_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f751ccd9e906a623_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.39, 0.05, 0.53], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cc2853d8b6012d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.1, 0.18, 0.39, 0.2, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7751c6f0cee1c777_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.08, 0.17, 0.45, 0.21, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7899499ff84a8e85_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74341e85d0177cb8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.53, 0.2], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958b058965c1b030_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26ce5db2672f4f7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.01, 0.05, 0.61, 0.25], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cda5e0ac04c178a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.1, 0.18, 0.4, 0.15, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3fcb5b04b27f083_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e86900e8b47c30d7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.44, 0.44, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a9b344c3038a211_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.72, 0.07], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16635cae28fefb23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.75, 0.09], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da7ec481091c60ab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.11, 0.4, 0.38, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e9bf19d0ef72f20_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_322cc44ab44ce591_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cbf32fc7d1357ee9_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9782ff5b193e3b74_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81e95ae83319c565_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.49, 0.24, 0.04, 0.23], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ab0765156c68db9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.07, 0.7, 0.2], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f36243997e53b9d4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.11, 0.49, 0.32, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7a9557c57aaf356_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64d7d5879157dd9d_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7efc135282c7cdd6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.02, 0.95], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5c84133c8316ac5_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc463a620cf009e5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.79, 0.12, 0.09], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c189c8ab38de005_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.4, 0.44, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69ce1c7f0f294e09_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.45, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_928e62f67aa8982b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.45, 0.46, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_975f0198d7d360bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.73, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44b186cbc8a235b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c2b55696ceaf618_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.76, 0.21, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abe5736ee767076d_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3c7a78d9e8bb81d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.23, 0.72], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a696407522a9bc64_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.08, 0.17, 0.45, 0.23, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43305fee0c96062c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.46, 0.4, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23557d20e60584ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffe1ec62c293df22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.17, 0.32, 0.03, 0.48], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85361e9b309ae20e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.23, 0.21, 0.37, 0.07, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_344d617d1921e28a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.85, 0.09], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85a64e89a32e5ad0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.5, 0.41, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f0d8dd4b2a39e29_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.9, 0.07], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7939721339970e4c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.05, 0.82, 0.1], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f6795aefb4a59e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da366145260381ac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.73, 0.12, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c17f0d955e8c034e_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d8e9374eeb0c220_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab798e52d496dd5e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.16, 0.08], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_299001dabd376d38_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.72, 0.17, 0.08], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e6f3adfea4b921_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d48806ad6db99c5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.07, 0.21, 0.43, 0.17, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c04838e8b5fb06ef_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48ae61237e67d20f_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b2b61ab647052f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfe52f0bbaa41d4e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.17, 0.22, 0.4, 0.12, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42a9187416b77eb3_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08bd68d7df82aed0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.57, 0.28, 0.09], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffd5a1454e5e09b8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.38, 0.57, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9aa654dbc67977a2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.24, 0.05, 0.01, 0.7], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_316db6800b7d00c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed6f4e25a786ceba_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25a498a7bc050295_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.79, 0.16, 0.04], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed04324ea2f1723a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.26, 0.65, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_658709dcd9250ee7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.77, 0.07, 0.01, 0.15], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1531e3b7afb2e3c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3feee735ee025de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.39, 0.52, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51f2fa5f7db0b5ba_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.64, 0.16, 0.1, 0.1], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee564c3af3d6e723_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.73, 0.23], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4547e0fc6bd827a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.34, 0.27, 0.02, 0.37], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43ee37918ac34570_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.13, 0.42, 0.44, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c7fc57e9d354d06_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.05, 0.91], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_217125549d0ef849_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.39, 0.49, 0.02], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ce599ca9f636b7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65be83247834569b_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c42a2dd75c512354_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.02, 0.06, 0.49, 0.33, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2ac8c87d695ed40_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.21, 0.08, 0.31, 0.2, 0.1], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8fb4fb48c4ae587_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.31, 0.04], "state": "An agent plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4dc7f9d70144ab55_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e0324ad81daffd_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fcc08a5f1e7e3fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.44, 0.32, 0.02], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d683c63273bdec79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.31, 0.1, 0.41], "state": "An agent plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd68627f81e01e73_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3753817c96e6034b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.58, 0.33, 0.08], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08e19f869b811d1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6ff3adf6afd9932_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.82, 0.16, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad2e3553df598ce4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.04, 0.14, 0.79], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_333ef51e959663e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0707, 0.2828, 0.4646, 0.1515, 0.0101], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c49845a9336ec5bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db682099776be57a_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bab324bde97e7e53_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.05, 0.29, 0.55, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18f9cdb9edf7a478_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.4, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5235c825e7da5452_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c434137389b39363_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.16, 0.42, 0.26, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09ede11f5a8fa5bb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.26, 0.64, 0.1], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a77402c521accad_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.45, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3efd336e433c2aa2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.62, 0.24, 0.08, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_891b76fb5739a4e1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.75, 0.19, 0.06], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aed1c184fa07bb3e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.89, 0.01, 0.03, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c629bc52c4d60e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ace44dc492a2e493_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee42e7dc67b9d591_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e28a6d2c0ade2843_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.31, 0.63, 0.03], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_652b46a3c48d6ba5_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a82bf8711f7925b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c73ece668808d341_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e704a65f6f83cfec_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95bbd461edd2200b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.58, 0.22, 0.14], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97a160edf1344a12_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.59, 0.27, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32e1d0bd2612c231_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.23, 0.29], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_514726ecff9f2185_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.45, 0.4, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40ef1e728335f865_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.45, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_854019f6b3e8f4f8_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8bf5defe7106c2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.03, 0.97], "state": "A task decomposition calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06b66891b017e435_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.41, 0.5, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e54371ca4e877a5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.75, 0.14, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d652fe35fc8fb541_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17bc3c1b8b5af404_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.14, 0.15, 0.46, 0.11, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e41fc0aee18f1c5a_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90b951b396fe70c9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.14, 0.53, 0.23, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6a43b48e74e2fde_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d498cb39ce5f3e34_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90731cb2152555ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fba3adb70e0e150_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.3, 0.56, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7628f9945a417fc8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.7, 0.07, 0.18], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c6542da9aae1377_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.27, 0.35, 0.3, 0.08], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c4f8af6914355d83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.4, 0.54, 0.02], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36356e38be922cb7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.11, 0.16, 0.34, 0.22, 0.08], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb8f6c056dcd1485_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ca8a692922bb9ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.09, 0.64, 0.2], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_726ca9d0b4cddf95_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d086f0b9364be664_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.32, 0.54, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_075b8790de6f9fbe_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9c96977ff33f5a4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0707, 0.1919, 0.6869, 0.0505], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b8a795515606e2a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.28, 0.62, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc2e564303995ea0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.76, 0.17, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af9273905b1d6f59_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3bbb8f61ba3727a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.25, 0.48, 0.14], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2677a0192ebb72d1_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d4bd6730b08629e_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36b736b87b466f18_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.09, 0.55, 0.34, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_200b82b4723e8185_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0dc56ca0a4e15a6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.3, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0002c468edcbd614_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.46, 0.34, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8264cc8dcae09ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.4, 0.42, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2d59e9a56cbd568_n", "kind": "noul", "options": ["false", "true"], "target": [0.32, 0.68], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3a8b8b6d0cb4aa4_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f3b816c3025bcb0_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0c127818b7718ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_848d9d0dab8a56af_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af0007c0abde98fd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.47, 0.41, 0.1], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba1e862ebb5fa056_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b567866ab85b9b9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.42, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e674e3e7840ab7e_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f45ac83d4a02031_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7beb6c6172b153ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_831144c32aca11c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45c318e75134c695_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.12, 0.34, 0.17, 0.37], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e63ddfd026f5683_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.55, 0.3, 0.14], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a151a60a891bd4f_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b025377bb7ed6ede_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.09, 0.86], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61e4e06bda6333c_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_579e8bb0bb2860d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.17, 0.47, 0.24, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab6934e360225844_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.44, 0.27], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e968dc3cb4f4db4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.23, 0.59, 0.13, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e22beddea8e9179f_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a60a14880b05b56_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.12, 0.15, 0.4, 0.17, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca2b45eb62a0d27_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.08, 0.37, 0.51, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0279ad5eadd45b1b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.15, 0.72, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4417a7bc1345cd6d_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd0441e3b09077c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1d23118b7ba63ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35f2fcf7a2deb8e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ebf2aaef76c9104_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.52, 0.45, 0.01], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75ba0b4562a1c07c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.43, 0.43, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d14e6d5fbcd6dfa0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.09, 0.91, 0.0], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b655876bb4420c1b_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02ddcd5932212adf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17ad7138867e11b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce8d2b8077eb5e54_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_385ac11f5cb83867_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.28, 0.61, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3bf291fc55286efa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.27, 0.25, 0.37], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c346fd85b3394058_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.06, 0.76, 0.17], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_776b5d10991808a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e6f3adfea4b921_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.44, 0.09], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87bed3086254f7f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d831c6f32694712_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5100fac467897a4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.31, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3a7106f06421323_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.52, 0.39, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87071155e73162a4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.15, 0.82], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9507f4ee6dab89e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.22, 0.62, 0.07], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abf6a0508dde06e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_659825ab9841cd1f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.36, 0.53, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caa64df60ad7a120_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1caddd02c110239d_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f96d485ce9b29d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.37, 0.17, 0.41], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd20aed82e23d0f4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.32, 0.62, 0.02], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd3078beedb424ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.35, 0.52, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415f43981ccdcae3_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05219f41b5403ab2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.1, 0.18, 0.47, 0.18, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec2927ab3f6f40f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.19, 0.16, 0.47, 0.09, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e123a4fc4a6acee_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee55b6d73875ff61_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.17, 0.33], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f47db1c0d1769609_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.38, 0.49, 0.09], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a850d0a86d295d1d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.51, 0.41, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44453df6927eec56_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.18, 0.79], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5799aa76d2d52fcf_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3a8b8b6d0cb4aa4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.3, 0.49, 0.14, 0.01], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b13a34da37bf8c34_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dddcb0dd5d8ad5f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.2, 0.74, 0.05], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ff922a778d8e64e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.17, 0.38, 0.29, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d87fdbdc1a2c90f_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf99b9060c7d84ae_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.34, 0.4, 0.07, 0.19], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_100bbc41f412975b_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc656df096979d57_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.11, 0.85], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c86c47d63aedd3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c278ccf5b704ad23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.57, 0.23, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17a647b6bf2687b2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.55, 0.23, 0.18], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fd1e0ac8834f163_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.23, 0.76, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fa9730d25309352_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.58, 0.13, 0.23], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc613c6d0aaf66e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.45, 0.4, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99f8bfca34dc59ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c2e042c1af35f52_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5847daf89d45149_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.0, 0.01, 0.19, 0.7, 0.07], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0705f5eff2ca22e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.45, 0.13], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2fbf3499ed1dc4b_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_151a6092be54c88b_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e9169a7b5ecb796_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93c30bef3767a53c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.3, 0.68, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd6ef8b42071c0fc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.56, 0.14, 0.27], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a48ac86858e07cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6102ac275a637c01_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9ef6c4e60169474_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a90ad2af4fd692e0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.37, 0.07, 0.24, 0.32], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e071f06d236c9a9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.38, 0.29, 0.06, 0.27], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df056231a7661c9e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.12, 0.72, 0.08], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8025f137d326edb7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.08, 0.72, 0.18], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3d8d3a99c61c671_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.08, 0.85, 0.07], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed3aad3d7c88001c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.1414, 0.8081, 0.0505], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8380e3eca8d53236_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.1, 0.19, 0.44, 0.19, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e24e2f5ea3157d6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.93, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1690638dd0771abe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.48, 0.09, 0.4, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77cabe14e8c36052_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e6d1bfeed21de41_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.62, 0.0], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_700ecd50bb38daa8_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fa2caa67bf70886_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.41, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff9aa6673204f93_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a8d9df0c7ecc172_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.34, 0.07], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b81f9891030cf43_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.27, 0.64, 0.05], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4513ebc3a2e4fada_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.61, 0.16, 0.18], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_506aacd399e08900_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38862ae2ba1f7e57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.09, 0.47, 0.33, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee0b0410493e1cc4_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b2124149073c7f2_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebee906dd88dc79c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.39, 0.09, 0.25, 0.08, 0.04], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b7118f62b93def1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.2, 0.2, 0.33, 0.14, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a0826e48f4947f8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.05, 0.19, 0.48, 0.21, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4cfbab243389aef_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42eec61568828ed3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.06, 0.45, 0.43, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9caf04a72ea8393_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.05, 0.93, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb69727c2c43faec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.16, 0.73, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54d5fa1b307e9820_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f615207b8ac41f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.31, 0.59, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a154c1c7866604de_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04ef04516d22c38b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.6, 0.09], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_937788fab522a730_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0b21bfc4172c472_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94e48928ed68a41c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.55, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68a90a47f083c553_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.47, 0.43, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d271b18350ba6b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.19, 0.11, 0.35, 0.35], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49de0dbf0922af98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.7, 0.1, 0.18, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17854a82cbf7ce39_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.29, 0.52, 0.07], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1c4766d6a22c512_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.87, 0.02, 0.06, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8cefde14f961d9e_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f970fb3f95896efe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.86, 0.11, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d03a7eb0d82e64e_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70e2b02436f82194_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.1, 0.17, 0.42, 0.18, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26d87b320661dc8f_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c61f473f4d7dfa4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.17, 0.14, 0.51], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba3afe1ec466d834_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2dd1df6151e9a4c5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.26, 0.68, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70338ec08652708e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.04, 0.35, 0.51, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b73a2f1b5608ff5d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.4, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_176b4cc1beae9130_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.42, 0.12, 0.28, 0.06, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_135f3b0f5b9a210b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.81, 0.05, 0.08, 0.06], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a36bb7a7287eeb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.17, 0.32, 0.44, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2583d2e88741dc32_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.08, 0.5, 0.34, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_601339f43d2072fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0961dc21d567c8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.48, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_164d98dc77209ab7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.2, 0.35, 0.39], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_77417bed6cf76430_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.42, 0.44, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_576083f6f0f02f51_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_592aec12475ac01e_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5f92bbf2bd41b14_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.31, 0.62, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1726f6daf1e8381_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b78a327f2b930281_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.29, 0.05, 0.64], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e5085bd7582db3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.86, 0.01, 0.09, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f0414f57d29acf1_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a98790343f971b1_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8a27da955a31497_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.07, 0.26, 0.58], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38f129a707ce886c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.42, 0.2], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5756a2510700b749_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7feeded03d46b560_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b3eabf7b7235eee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.06, 0.43, 0.41, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5209a026db1a622b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.11, 0.16, 0.44, 0.22, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06a2907b594402dc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.32, 0.11, 0.28], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a60d7f77eaab32bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.31, 0.18, 0.06, 0.45], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e23d3ed5c4fb65f_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6fcab09f8d51e2f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.36, 0.09, 0.23, 0.13, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba75a7a8ec9bc9b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.03, 0.14, 0.5, 0.24, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_160b1de8d8ec42f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2292abe1e27d7b14_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_898f63d74b0372ae_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.6, 0.3, 0.09], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_136c6e0264e153e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.08, 0.17, 0.39, 0.2, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53e11686442fc94f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.73, 0.08, 0.17], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7093b3469fdc55d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.44, 0.04, 0.19, 0.07, 0.06], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8a76aedf569fbbf_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe93e33a6d486c37_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b915dfdbdf3f7e90_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.09, 0.82, 0.08], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5bf992aea4e38392_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.38, 0.52, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b471ea2014e4380e_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce18f8e67c266d79_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_982a3f76b092803a_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af4dd96e411d8c26_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.28, 0.58, 0.06], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46e25b4310197415_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.27, 0.35, 0.32], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44c3a3df716fd00b_n", "kind": "noul", "options": ["false", "true"], "target": [0.23, 0.77], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c19f0943141555f_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_904aea898f633d63_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.53, 0.39, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3753817c96e6034b_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22a69990af1c557f_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fe400bacdd7c19c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3232, 0.5657, 0.1111], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca4a95e0ab35d83c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.66, 0.13, 0.2], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a4c3fc57e37a4e3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.58, 0.15], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fbcb8b0410fdc6a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.11, 0.45, 0.33, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4920a534048d41aa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.1, 0.2, 0.42, 0.17, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e00a6948ac35f8c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8b4a67704e3ad6c_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f84f1ad928eae97e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.2, 0.14], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c20b6926910828f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.19, 0.46, 0.27, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29968b71c569ff6c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.33, 0.13, 0.3, 0.1, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_497c75b79da47a75_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10e7d1c544d628fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8182, 0.1717, 0.0101], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_361dce359ea41dfa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.36, 0.56, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3209b7712c92c0be_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.95, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f970fb3f95896efe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.44, 0.49, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8309b3145c160337_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0101, 0.0808, 0.5657, 0.3333, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_250b9a909f2c78c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bbffd9e161e322c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.79, 0.14, 0.07], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_885888423eb70e74_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.11, 0.38, 0.42, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_848e67ded2abc84e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.18, 0.18, 0.42, 0.12, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac758c61b81e011a_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1b7daf75bad78c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51bdf571be3b5933_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d6087efb0ce8494_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.29, 0.6, 0.09, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fa17a4512fca752_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.03, 0.95, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63c5fb84521bb8a8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.43, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e28a6d2c0ade2843_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ea8cfebfdc6919_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_742a70b645503ecf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.01, 0.08, 0.81], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d8e9374eeb0c220_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.51, 0.15], "state": "An agent plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eda057ce1d539889_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.17, 0.16, 0.41, 0.11, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e82e732f93e60dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.12, 0.45, 0.32, 0.02], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85511f46a92de3ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf4d0d885a7a4393_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.81, 0.15], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac2e7a724d02fed8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.45, 0.08, 0.01, 0.46], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a6321fed1eb5486_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.38, 0.47, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85d5520848570379_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.79, 0.1], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5757f1b17f3c81_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d07ac07f4b544a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d7992c70c9b9475_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.06, 0.21, 0.63, 0.1, 0.0], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58fb3ead4216207e_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40fc447ea5271f26_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.72, 0.03], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_500e8326fb40eb56_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.56, 0.11, 0.04, 0.29], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40836617b69dbae4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.22, 0.09, 0.65], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a510160b5e3d3d88_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.53, 0.06, 0.27, 0.14], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_458f2468a2ebf546_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a63899ffbc214fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.4, 0.44, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fe400bacdd7c19c_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1a35edc9ef07baa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.11, 0.28, 0.52, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da7b7735c3d6707b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.09, 0.48, 0.33, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6c664550416bde1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.14, 0.51, 0.21, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a948ed37dd97540b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.61, 0.05], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_596e0baf008eb976_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.34, 0.5, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6f3e0eaaae957d5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.02, 0.44, 0.42], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba24eb7453ad25eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.29, 0.59, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54c97a589689f9f1_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An automation plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc5c89de83b26ed5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.32, 0.5, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3684caf19a02cec4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.38, 0.6, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1f0e06014104a45_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aae0d29dcf2f08ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.19, 0.22, 0.37, 0.12, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1409082087b1dfd7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.19, 0.65, 0.1], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ce6e4fb5d3fd39b_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29a8c5dc914f6ad6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.05, 0.54, 0.38], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efb1ba9965dd9a29_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.21, 0.2, 0.38, 0.1, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_843cc9bce04020fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7384579268b43d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.48, 0.03], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18428a8347470f38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.19, 0.42, 0.26, 0.02], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35b143a9f03a4cf6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.01, 0.04, 0.43, 0.44], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff3359e314da8553_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.52, 0.4, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_489b4b3596340e77_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457b7ea1440e0296_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8accae430ea8b54d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.23, 0.13, 0.35, 0.17, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_07040ed65fd98c14_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f5df01eb8b81488_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.19, 0.56, 0.23, 0.01], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7af3356850512d7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d79e03b47aa640_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ec701efac4c37b0_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65994e45b04af382_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.24, 0.51, 0.14, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f08b80af0d4a7c7e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.14, 0.48, 0.32, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76825080c8b87f48_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.02, 0.94], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a6321fed1eb5486_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3f26a340d4defa37_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a5a9421a2ad5d9b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.54, 0.06, 0.38], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_838f852d34e0be1a_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b4dac0b8f1fce94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.4, 0.47, 0.03], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1268d0b86551ff3c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.34, 0.61, 0.02], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95f585062ae50219_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c96233c729a9e5e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.06, 0.84, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3229945a7f8469e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.41, 0.5, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c2de955856f4611_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7058def01b14c20_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26535e9f82f0557_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_546701b9a7d0a50c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.75, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d388da3704f721_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.57, 0.34, 0.08], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_378109ff13540160_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.4, 0.52, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a33d8d05eeaffa4_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76628ec57c756596_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.2, 0.63, 0.1], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45731559be6b926d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.01, 0.02, 0.1, 0.58, 0.14], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05e965e825eb38e8_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_188889171adfd918_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.16, 0.7, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757e864169e8f701_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.39, 0.04], "state": "A task decomposition requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_437894c57f111804_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92ecefa8b68abc93_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_374110c228179029_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.39, 0.53, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c96ee0c9f61362b1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.19, 0.48, 0.3, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21ef33cbaf481fc8_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f9230f1146f55ef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.15, 0.42, 0.28, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_530d52bca050c182_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fb5ac91f58dcfd8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.43, 0.43, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee0b0410493e1cc4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.1, 0.9, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55bb6351f9613d89_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.47, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3e6eed98e0adad8_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e73cdd46cc5bafad_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06a2907b594402dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8886147699a1f3ef_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.77, 0.03], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae93774209e1d29f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.38, 0.51, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_126da9e839b3f1c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cd8910b3526d8e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.78, 0.22, 0.0], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80e9fc8766359cf8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.5, 0.44, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39919b1ff3e12e7e_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06a2907b594402dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.24, 0.19, 0.4, 0.09, 0.01], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e43c31528acbc00_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.35, 0.07], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_530d52bca050c182_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0101, 0.0404, 0.3232, 0.5657, 0.0303], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a11c67c14555a9d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.37, 0.47, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_236be50aa8a0bd05_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.55, 0.04], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1aacb2f3f0fc7028_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.34, 0.56, 0.08], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa9f468c44113d6f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.5, 0.2, 0.17], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757d084c78c7a495_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ae67422f28fa5a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.09, 0.58, 0.26], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8380e3eca8d53236_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.23, 0.62, 0.13], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b255d7a2a2b38490_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.37, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_592aec12475ac01e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.22, 0.21, 0.32, 0.14, 0.05], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74a45807872fc87a_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a10bf0f4c787697_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.21, 0.45, 0.24, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e80bfd1d34c54e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.36, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce5f7b177feace42_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.09, 0.85, 0.06], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8309b3145c160337_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fcc88289efa28377_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1c49d95821acf91_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.15, 0.62, 0.19, 0.01], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_164d98dc77209ab7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.32, 0.51, 0.12, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b1f4d068952d6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.25, 0.21, 0.33, 0.07, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29e1dbe1671bdd21_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.64, 0.2, 0.14], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74a45807872fc87a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.06, 0.22, 0.53, 0.1], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415f43981ccdcae3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.88, 0.04, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1955980e8928d16_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.21, 0.13, 0.37, 0.15, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_910ceb25f05a8dbb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.16, 0.24, 0.44, 0.08, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5dc6b384a91f22a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.04, 0.11, 0.45, 0.29, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57c9825def0e7a2a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.15, 0.21, 0.39, 0.16, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c708c9f85e4a782_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.52, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acbfbb8c105d2748_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.79, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf5bf542f98895cc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.36, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f42a9965997ac00_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.01, 0.54, 0.4], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b778c1eb9e0804cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45c318e75134c695_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f091b0967ce7ff53_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb56c288cd8d34ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.19, 0.73, 0.06, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85361e9b309ae20e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.4, 0.18, 0.37], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f331226ae4d27b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.1, 0.18, 0.5, 0.12, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_700ecd50bb38daa8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.24, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca4a95e0ab35d83c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.16, 0.46, 0.28, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a979b919d1d3034_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.22, 0.38, 0.16, 0.24], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3fcb5b04b27f083_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.19, 0.65, 0.13, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8933a7d560cbc9e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.53, 0.37, 0.08], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4dc7f9d70144ab55_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.58, 0.06, 0.04, 0.32], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ababd64a24d94af4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.62, 0.02, 0.24, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c3597d810b8a9bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.16, 0.48, 0.32, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e330d12a895ef8a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.56, 0.05], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d06fd2a0509d26f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f89c97182a70660_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.1, 0.06, 0.75], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_334a63b538362a06_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.13, 0.7, 0.1], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3eeffff5a9955317_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4761fe14f277aab3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.45, 0.45, 0.08], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57791e80ef540d39_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff7ec5ffebc09d57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.08, 0.16, 0.41, 0.24, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db17f02970a5ffce_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e60b934ebfe6cab8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.2, 0.71, 0.05], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ac9e32068f182f5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.92, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caefc75f9f492970_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.24, 0.2, 0.37, 0.09, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457efabe8101eaa2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.21, 0.12, 0.38, 0.12, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ca067c8d7da4c0b_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71ca1a4fa2bc397e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.47, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8fb4fb48c4ae587_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.74, 0.12], "state": "An agent plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_731695004796ed3b_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_816891706439634a_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a57547febd908917_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.13, 0.17], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04f8e4c0aa4dbdb7_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5507ee782f2e46a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.11, 0.13, 0.44, 0.16, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69fd5a212438894f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.22, 0.05, 0.21, 0.52], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69ef1e2607449942_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.84, 0.01, 0.08, 0.07], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0100824433beb91a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.13, 0.46, 0.32, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35782e8cb2da671b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.2, 0.09, 0.08, 0.63], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c2fbe5cd2f83545_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.6, 0.13, 0.19], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfefec13b8322d6d_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9368c35590e1c89a_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1890a59accfe573_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.63, 0.1], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8809a18ce1d14791_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.43, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d03a7eb0d82e64e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.1, 0.29, 0.47, 0.05], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b13a34da37bf8c34_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_011c67be20284c41_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.32, 0.03], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc60bb7eb2af6aff_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.08, 0.8, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94e076f83aee26bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1dbc7b4473888c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da9d31fdc7bb7e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.2, 0.22, 0.39, 0.08, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bbdf0505e4a71a1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.48, 0.33, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3169b463c94a8358_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.16, 0.17, 0.35, 0.16, 0.04], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8319342981a9a54_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a2e024d83329fff_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.48, 0.07], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66ac27615660abda_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc934995de493385_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_417ae5dc0115cf4b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.28, 0.44, 0.16, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd0d86ebb49f4451_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5958de6facfddd23_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.6, 0.15, 0.22], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff4e779b8d4f50c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.16, 0.46, 0.22, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63559a1ecb211141_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.78, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18000b5f7aa7ec98_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6a044dcf0377ef5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.15, 0.57, 0.18, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2a39846d1f2a3eb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.08, 0.45, 0.31], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da744e7b59d313e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.3, 0.61, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed98613e6f982c13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.2, 0.16, 0.45, 0.09, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_feba41da31699030_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.16, 0.83, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_886a66b1b1ffa27e_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e9e86be4b6f5a83_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.29, 0.34], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df7011e46a340e3c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.21, 0.35, 0.43, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b4fdee68bf5ccbc_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ac721995d72461c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.17, 0.44, 0.25, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c34d33694f626a94_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.07, 0.17], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7bfc494df59b5cc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.62, 0.09, 0.26], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab73448bd1f6f96a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.6, 0.24, 0.15], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67b0a9daa285081b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.25, 0.59, 0.07], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6579746b10a4644_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_278e395e631fd4a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.3, 0.14, 0.35, 0.1, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be4c0d9b40bef977_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5770a524b2afc506_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.51, 0.29, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d921cc5a6ba9c4a9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.43, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04bb0a061b179071_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57918d94c7408ef4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.36, 0.57, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d13601f00d11c09_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2bcde3b10834015_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.17, 0.2, 0.38, 0.14, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1896a282d360f25a_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e0324ad81daffd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.41, 0.45, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a34f40cb83cc6e44_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e206d508c217e03_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.44, 0.1, 0.43], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9e961fb3ed33e3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8accae430ea8b54d_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfefec13b8322d6d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.3, 0.15], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98f655f782787f37_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.65, 0.32, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c50ff7b399fb09e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e982e1d61c1a431_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.06, 0.2, 0.54, 0.13], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37b9b95bd48e8abe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0808, 0.0202, 0.0404, 0.2121, 0.5657, 0.0808], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f3b816c3025bcb0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.04, 0.23, 0.55, 0.06], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd96ba7d24fb6302_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f79ad6dc79db39d3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.06, 0.77, 0.12], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_028039e1e25fcd15_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.45, 0.32, 0.03, 0.2], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_339bf2f73c6e149e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.36, 0.25, 0.11, 0.28], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_738eb66f1f2166ee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.05, 0.78, 0.17], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2a82438b5e028cd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.48, 0.33, 0.02], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3143aa8812ab15c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.3, 0.14], "state": "An agent plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d336a11db32d032_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.23, 0.76, 0.0], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e96b6b16b7999f38_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c0397515dae2b5b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.63, 0.1], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca18614b459f3e6_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5ecdcf36665ade4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.11, 0.2, 0.35, 0.24, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caef6e13019866f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5eae093b10c6d286_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.68, 0.14, 0.12], "state": "An agent plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4e99dde9304253b_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942ac662dd6f9bf2_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c20e8f9be4102b94_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53fadb4a6aa3162d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.49, 0.06, 0.23, 0.22], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee4559b273e3531f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.13, 0.67, 0.14], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3ebd40a14cfa4e2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.83, 0.1, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20a969f6e9f1e409_n", "kind": "noul", "options": ["false", "true"], "target": [0.22, 0.78], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c26a518b1f108c42_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.02, 0.09, 0.15], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a73ac89663d2aae_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02cea0704ecedd4b_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3da18e91917ce4c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_335ac00b28667867_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.19, 0.28, 0.35, 0.09, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f23b236f960605c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_897cf18127c0ab61_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.06, 0.85, 0.08], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c025d60ed07af741_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abea6d3f20a042c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3d2880c330d4177_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f94e1158f0d9a6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6ff94256d3d1abd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.49, 0.24, 0.04, 0.23], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6933df6e028c48e_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_203005c5e6304c45_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.17, 0.54, 0.2, 0.09], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9dad05700ff379f_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e4cf06d1de55e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.13, 0.57, 0.21], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4f8a800dc49aca1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.75, 0.04], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8eb765f46752ad1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.07, 0.43, 0.38, 0.02], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7af3356850512d7c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.11, 0.18, 0.36, 0.22, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1faaccacdf15ab05_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f170fe50b0271eb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.28, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d383dcd0814b7642_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98b1ea440d746490_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.21, 0.65, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6a13b869121a77_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.03, 0.33, 0.48, 0.15, 0.01], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3ae8d9042fd2447_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eaa209b839a4a32a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.41, 0.02], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a9bf3f506759baf_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f148582b365c909_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.17, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b21b3c3bbeb5bb9c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.47, 0.1], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_baeea12bf2d1c921_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.95, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3087b083a7ca9c59_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.37, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1f5b7a8156f66d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.15, 0.58, 0.26, 0.0], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcc6dd27f1235653_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b949182670528683_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e38911b976da5cf6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.48, 0.13], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f17549934649ab44_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.53, 0.13], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7a2244760cf3c3f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.09, 0.32, 0.41, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24f1643ba878b640_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aeaa72abfe290c5e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.54, 0.3, 0.15], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99605f6131a71ec9_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ce80ec26abd5068_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.44, 0.39, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84fb23aa0100ef2b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.67, 0.06], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70b3fc788b4930ca_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.54, 0.1], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ab68c85a6b1c382_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.37, 0.41, 0.21], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c8cc2539136eeb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_716cf00be3013d50_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.17, 0.65, 0.17], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d5e5fb26f04e12e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.18, 0.44, 0.35, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d3b3a97122124e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.05, 0.7, 0.25, 0.0], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e70d1f7cda39b5d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.07, 0.38, 0.51, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8304af1aab219eda_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba75a7a8ec9bc9b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e36ad73d73247b0e_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_062db51b690dc390_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.38, 0.47, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0ebf81b4c4254a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.1, 0.33, 0.43, 0.09, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_adf131cb77662c46_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.1, 0.18, 0.44, 0.14, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_321487005fce5a72_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.1, 0.44, 0.32, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed8447c104615b87_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.14, 0.77, 0.06], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b915dfdbdf3f7e90_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d0d683794af532d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.58, 0.2, 0.04, 0.18], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c92e0c387f352af_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.11, 0.73, 0.12], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3087b083a7ca9c59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0505, 0.3232, 0.5455, 0.0505], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4547e0fc6bd827a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.21, 0.18, 0.37, 0.12, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93ca9de9da036107_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f331226ae4d27b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.49, 0.07, 0.25], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3978ec1143b544e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.25, 0.13, 0.41, 0.1, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664137f0f040ed34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.09, 0.19, 0.41, 0.25, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73e4b79db552ef22_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8858b20ce5f77ec7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.55, 0.24, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88313f9c874fa8b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_528bb034edcc314e_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3760299571e93537_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.06, 0.7, 0.21], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef8a8463a44fd603_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.86, 0.13, 0.01], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a7d9cd7eb64c642_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.01, 0.03, 0.6], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21acd689e4f191da_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.61, 0.15, 0.1], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edd61b957f1466d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.11, 0.22, 0.43, 0.16, 0.03], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b5eff964580e606_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2a3caa99a4fe6ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.2, 0.51, 0.11, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea2d938aeec9c4b1_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5237bc2854fddbd4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.05, 0.92], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415da395b7b74348_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e9be74f53c3e129_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.16, 0.32, 0.39], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb169bccf2c60e39_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0e022fb3d28d283_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.23, 0.6, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a98790343f971b1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.51, 0.22, 0.21], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16635cae28fefb23_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.47, 0.16, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1217cc4e8c6381e_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fbcb8b0410fdc6a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.37, 0.09], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8b4a67704e3ad6c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.25, 0.64, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dda10c06e3758286_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.19, 0.65, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fede1f7541a5779f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.23, 0.62, 0.06], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cdfad2ec70a5c11_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.04, 0.73, 0.22], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1c3e2fb1d4a683a_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d6491d630e354c8_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9932ac390479b8db_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1140993bdd06b32_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.61, 0.35, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab4e54aaacf5bd2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.23, 0.49, 0.16, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_126da9e839b3f1c3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.27, 0.29, 0.26], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ad935d8218c79cf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.19, 0.36, 0.35], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb8cff9477c0ab46_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.45, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_075b8790de6f9fbe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.49, 0.4, 0.02], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28fa653b2eb7b43c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.07, 0.78, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfecae896770d57f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.38, 0.5, 0.02], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da087aadb2701e6e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.4, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c07c8809b00a3dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9dbba3b96ecb1d2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26ce5db2672f4f7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_623cf5d202140573_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.53, 0.1], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81039c3b81bad20f_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c1cd571d2707e14_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.42, 0.25], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e38609392eeea46_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_532fc42cf951190b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.05, 0.78, 0.15], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c8d5ad73f0e75ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b993ff2a040bfc64_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.44, 0.37, 0.06, 0.13], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5de1114bd9c12cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e78594bfb404a804_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a059ba9c436ff76_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c350bea5065775e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f77a9a22fb2c3429_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.61, 0.15], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5dd56afa77554dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.05, 0.72, 0.22], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fff5f2504c696101_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.02, 0.57, 0.41], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a7b77027cd5b1f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0303, 0.0202, 0.1212, 0.5455, 0.2727, 0.0101], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_899a1381f91590d6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.6, 0.11], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ce80ec26abd5068_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.13, 0.75, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9338336816543274_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ec9e05f6d541467_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c84b178eed9bb32_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.17, 0.16, 0.46, 0.13, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c07c8809b00a3dc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.34, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98f655f782787f37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.22, 0.59, 0.1], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31b46eef4935ec18_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4082dfb31a3f74e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.62, 0.14, 0.16], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_149225c65f9affd1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.19, 0.49, 0.22, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45f37b88cc4aef9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1dd511831a507dd5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1111, 0.0202, 0.0404, 0.2323, 0.5455, 0.0505], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37eb72145386aaac_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b8304f82804214e_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a10129fefd69e8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.49, 0.29, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c93c08e5ff47243_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15324ff852a46665_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.68, 0.21, 0.07], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19bfcec860ee3b1b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.52, 0.41, 0.02], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21fcf6e75dc00c2f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.16, 0.68, 0.09], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1850cf203d2597b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.32, 0.55, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f95babb6c57ab22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.16, 0.72, 0.09], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13c8abd3e6749b00_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.22, 0.63, 0.07], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc859c72bde181a0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.38, 0.53, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f352310972059e5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.2, 0.77, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d57848d82b7737b_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08a211b33848eb52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.13, 0.43, 0.31, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a032a2ba417d2ce6_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed3aad3d7c88001c_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe93e33a6d486c37_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.75, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4963decf96f76b75_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.1, 0.44, 0.35, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_677bbc368d055a9a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.03, 0.63, 0.32], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea4ea17d3b05666d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0505, 0.3333, 0.5354, 0.0505], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56f99c8bb8de7d43_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ded8c953aed34c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.17, 0.13, 0.44, 0.15, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eda7f78582fe0d1a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.24, 0.62, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2677a0192ebb72d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.49, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bbe20ef5ebe63435_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.68, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93814c9337a9b736_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dddaa76826ef1ebc_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9052bb24eff5a39f_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_30120f8726c4334c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.89, 0.1], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f0589a1b10f30e_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c11fdefaeee82fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4619a29ab6084ab0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.0, 0.09, 0.79, 0.1], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35041b6f6e91c569_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.45, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a86527f4379aaf48_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.54, 0.35, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3765236dc0454c22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.66, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d55c6dd59b69ddaa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.4, 0.29, 0.3], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_592823c8399070df_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fd119c481e316d3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.08, 0.38, 0.22], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9fcb43c43bc1151_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.54, 0.41, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9aad87276a489840_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.45, 0.44, 0.02], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51cb60461e068ac2_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1c49d95821acf91_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_58671bcb7cbc84fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.35, 0.06], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a4c3fc57e37a4e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.13, 0.65, 0.15], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac6285784d4a38c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ce599ca9f636b7c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.38, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17d4175415b5aff7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0101, 0.0303, 0.3434, 0.5455, 0.0202], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1faaccacdf15ab05_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.76, 0.22, 0.02], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aba5f3908086fc5_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e00a6948ac35f8c1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.27, 0.32, 0.17], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e123a4fc4a6acee_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2af4e72265d54d2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.14, 0.49, 0.27, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5173494e2c30f87_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_543a82e57d9c92c5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.09, 0.25, 0.44, 0.16, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d46b62e22f01c00_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.47, 0.38, 0.06], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23208506cd090c79_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66820962a5b2331_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.07, 0.14, 0.56, 0.19, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f0deb9bd988c07a_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66b4f5086c99afff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.23, 0.16, 0.32, 0.11, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96f77196de8f2658_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.25, 0.49, 0.17, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14b6d52777778be2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.75, 0.02, 0.14, 0.09], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc854bc563a88d90_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba334f417b47b8f0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.13, 0.38, 0.41, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca182b778ec05216_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.76, 0.2, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5aab4ab386f4a1a1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.28, 0.61, 0.04], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_182fbf9911ed2edf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.04, 0.95], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b7c88bbeaa866f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.35, 0.62, 0.02], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bab0f0817a36018c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.98, 0.0], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd476f30e5788c91_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96b1c32131d4b147_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.59, 0.38, 0.0], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a81efe2d6089cd13_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e554f090acc8542a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.13, 0.48, 0.23, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4da32cdbca57989_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c9f7b4481634fa6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.17, 0.43, 0.3, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3d928547602b222_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.32, 0.53, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0af2813c972afb3_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8509958fab1aeff3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.43, 0.53], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa5f2d01df0b2d14_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.3, 0.58, 0.04], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee55b6d73875ff61_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.48, 0.35, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d54013cfe6c280c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.39, 0.48, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc1e82e4ba61d760_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53948a31aee00e23_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be7fa7c56f2be55a_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_828cdeffa38dce37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.52, 0.27, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d45502fe97a8ed9_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9594c6d840194460_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_226bec6212fdabed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.04, 0.17, 0.57, 0.13], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_175128c07b0e1b97_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.28, 0.59, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ae4bb54cf117b33_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.04, 0.19, 0.61, 0.08], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bebd57a1b061ff9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.14, 0.59, 0.2, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8a4188f07f4d3ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.28, 0.1, 0.3, 0.12, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ac85f8a4234dedf_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d43940b6c9501535_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.29, 0.2, 0.47], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49faad74a96daf20_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.49, 0.37, 0.02], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af9273905b1d6f59_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.01, 0.5, 0.43], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0188f64b18f0d5c0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.23, 0.64, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72b2f1099320a361_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.25, 0.14, 0.33, 0.1, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66d2495125e6d86c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.63, 0.06, 0.25, 0.06], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ab7589464b00097_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.07, 0.33, 0.56, 0.03], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_463fd86d4d16bd42_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.33, 0.1, 0.31, 0.12, 0.07], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1d1706f3b1b73c6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.11, 0.88], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0a0489ce3686a4f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.28, 0.09], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bafa7e660d147f48_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a25336df8d6d89ff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.33, 0.64, 0.02], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da744e7b59d313e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_654a98c3602d6822_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.26, 0.01], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f86866d2979e760e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.11, 0.53, 0.23, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b41017e859db61aa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.37, 0.21, 0.26, 0.16], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e24c6c69d5d1de3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.26, 0.03, 0.01, 0.7], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d437029cd105254f_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_421705ea38b712d0_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_385ac11f5cb83867_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.61, 0.28, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11fd7eacb93f69c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.4, 0.09], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6594a21bf8de74dd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.81, 0.18, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce3f0d0ab4125653_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.15, 0.13, 0.39, 0.17, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b75b8cdbdc3be5f_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87b139db602915f8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.34, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b10fb9fddc3dc254_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1c62223a188078b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.19, 0.7, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61d6d3e30c30edbe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.23, 0.12], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4465e1914854dc9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.04, 0.26, 0.58], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eb33ee004eaddc34_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0127b64efe39e61_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.46, 0.47, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_78649bbe4e9d56e8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.4, 0.05, 0.39, 0.16], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96ec67e6700d5472_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.15, 0.3, 0.51, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c54dc84ffa7e8e1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.62, 0.26, 0.08], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1f5b7a8156f66d2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6667, 0.2828, 0.0505], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d6aa770e70f57ea_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e038ff0a6931779_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b057d38289de2ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c30825b85491e92_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.35, 0.6, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3bc981255183ba8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.0, 0.09, 0.83], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0af2813c972afb3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.34, 0.54, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855831e0df346ecf_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eb20552768160c34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.03, 0.18, 0.66, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a2a800ba6d09a32_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.35, 0.06, 0.56], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18ce66bc6c378fcb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.38, 0.5, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4280603538002c00_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.42, 0.25, 0.04, 0.29], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7140838b6cccc0a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0139a653370db058_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.15, 0.42, 0.36, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df7011e46a340e3c_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21f500b8b11a04d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.04, 0.6, 0.34], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ed52b0691fdd43e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.53, 0.14], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1353a18bc0141bd3_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34525ee4851c3efb_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69745af02b44ba3e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0202, 0.5657, 0.3333, 0.0808], "state": "An agent plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b18885ad84792403_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.65, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_559773473da5fd9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee4ae26bf8284342_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40405ee6417b3a11_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1414, 0.2828, 0.4646, 0.1111], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_442f16a6d9a0f15e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.29, 0.51, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85d5520848570379_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.65, 0.22, 0.07, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7298ae159b451fac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.73, 0.01, 0.18, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42c3eaf0ad0be25f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.59, 0.16, 0.24, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_179c9e5eb6d9d06a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.76, 0.19, 0.04], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96f33b72ef10a06c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.48, 0.38, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_011c67be20284c41_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fa2caa67bf70886_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.42, 0.46, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1306acffc48aefb0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.12, 0.39, 0.42, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf590a676c77fa39_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.14, 0.03, 0.8], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfe52f0bbaa41d4e_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cba376fa60c1709_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.13, 0.56, 0.21, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2fbf3499ed1dc4b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.53, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5efcf89cd0f623d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.74, 0.09, 0.15], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb8cff9477c0ab46_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.17, 0.46, 0.33, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_006adc6f46720911_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68c4bb1c1eaa6238_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.26, 0.6, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420932e343b3f53f_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d6b667a8d0e2e6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.52, 0.43, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d8075e8bc272c29_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.62, 0.15, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41b0f7e09579be93_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.23, 0.16], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_adce541271349f46_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.11, 0.45, 0.36, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3da18e91917ce4c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.16, 0.24, 0.55], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2a40ec5dd180fd28_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b057d38289de2ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.15, 0.06], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10666fc2532c9e75_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.0, 0.97], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_743d50a4d742638c_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fe84315f1ee110a_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85327ca3b3928bfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.39, 0.5, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cca83317162ade41_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.31, 0.31, 0.24], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6563add88809eda7_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f148582b365c909_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f91d1dc39a1eedfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.26, 0.15, 0.38, 0.07, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1441e1a2a739762a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.38, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_554ddca1aed4c4a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99f8bfca34dc59ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.02, 0.25, 0.68, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce11ec06e01636c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.33, 0.54, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec15e95648c8753d_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc384c67f2e9dcf7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.33, 0.58, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec1f535871b5939d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.21, 0.31, 0.41, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc193c3d3bd5b5af_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f931558db8cd32ce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.17, 0.44, 0.26, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_127cee9cbff83a52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.07, 0.71, 0.19], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40836617b69dbae4_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_517bd82e24735363_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.05, 0.11, 0.49, 0.22, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0705f5eff2ca22e_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a715b907e4a223ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.05, 0.4, 0.53, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f028c40ab18d8ad5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.0, 0.0, 0.03, 0.51, 0.4], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3bc981255183ba8_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_adf131cb77662c46_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6acb6fa1116ac9f9_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_452723316aaed817_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff7ec5ffebc09d57_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.08, 0.86, 0.06], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0104a22dc61ed6a0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.01, 0.01, 0.62], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f745051580a558be_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.1, 0.83, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac9e89a5474bcd6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.16, 0.28, 0.41, 0.07, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9bca918a7877c21_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a510160b5e3d3d88_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0de791aa3609e92e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.47, 0.07], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_632cd17dd8f30632_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b767c931f026433e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.23, 0.73], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e48816ba5ec3910_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.77, 0.16, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bc8b24a2f29e037_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.11, 0.09, 0.71], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41fd94c0d9caad07_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.27, 0.61, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a57e2d0aff81f2f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.38, 0.53, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc2e564303995ea0_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_440a473d3c30c59a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.1, 0.18, 0.45, 0.21, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48b0e98a4a66b49c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.55, 0.42, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76089f4abe448de6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.65, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8199c5974489484e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.34, 0.43, 0.13], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cf925dc454c1e30_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.1, 0.52, 0.27, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_182fbf9911ed2edf_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_46234e5c609bfbfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c979871eafd72afb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.33, 0.07, 0.36, 0.24], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ae43a6a43f7f434_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a46c35fba63812d0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.61, 0.03, 0.01, 0.35], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12f86d3bcd3b66ab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.35, 0.55, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b7d45acfa2750a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.36, 0.51, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86cbef0dcae9ef66_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7237c523643d12d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1c62223a188078b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.61, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea87355cb7e78d5b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.2, 0.14, 0.65], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e97e2ad8dfc9749_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.22, 0.69, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86990f33ce63e7a0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.1, 0.87, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a81d184401dc47c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.1, 0.37, 0.46, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91506538ffecf3de_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.82, 0.15, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e28f9bd6ddadb14_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.07, 0.44, 0.44, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0efaf93756ab9f2d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.21, 0.42, 0.32, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6375e93f2f6ae957_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba75ca2ac4dcba72_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75e18403c29c782_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1e32d70bd91fd37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.3, 0.55, 0.05], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f3d1e3f81fdb7ea_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.26, 0.16, 0.55], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6071d1f8adfbb87_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.06, 0.57, 0.35, 0.01], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ad07669cfd1dd7e_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7e5c90b588d237f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.54, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f1609e8bc644fe7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.62, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5dfd0c3a1d1267b_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d9f1a7e976e1c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.4, 0.03, 0.38, 0.19], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f42a9965997ac00_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39fdeb6d5212dcd9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.07, 0.37, 0.47, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9af3f58eaa02118_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffe1ec62c293df22_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_845a76dd016805e3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.91, 0.01, 0.04, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02208119118eedd6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.53, 0.25, 0.17], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7e3b50d24648695_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d7d862a8042f4dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.05, 0.11, 0.45, 0.26, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855b343674aa5c6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.41, 0.49, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a425293af304c507_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.59, 0.33, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ef9ea52fd57c913_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.66, 0.04, 0.26, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f87e544e6935872_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.1, 0.56, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b93387ffcd5495e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.38, 0.28, 0.33], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebc6eb08cc13b72c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.19, 0.33, 0.06, 0.3, 0.07, 0.05], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c67ebfeff6c3464d_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea787f435ab126cc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.29, 0.13, 0.31, 0.09, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e39dd521aaba9bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.13, 0.75, 0.08], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02c9eef1af87fcc3_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c772aafca72c873_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.02, 0.12, 0.84], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1b0cc201477bd56_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df4853dcb77320ec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.08, 0.38, 0.51, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_65d1dc0ca3bc747c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.12, 0.47, 0.3, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab40858c20e66da9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.15, 0.47, 0.23, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f353d1ffafaa60ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_189c223d67565875_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d58a7fdc31dca408_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e004f033fb6d15_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a180bbc6105b776a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.2323, 0.202, 0.3333, 0.1414, 0.0404], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_834929b62c6a1bdf_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_558cb81d92c77671_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.59, 0.03, 0.29, 0.09], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dea890377f33e4ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7237e44ac0037f62_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37b280193138c49c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.56, 0.07], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dfa46782de59924_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.08, 0.39, 0.45, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1d1706f3b1b73c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0faaee4227814672_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.0, 0.07, 0.87], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5549ccef0870e06b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.6, 0.39, 0.0], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38d445f75b4e34f7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.92, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5dc6b384a91f22a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.29, 0.19, 0.47], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6fc617bfc967933_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.4, 0.39, 0.05, 0.16], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de44ea1663f30b7e_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3979472bd68dbc6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.38, 0.41, 0.18], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6246b31361574e6_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a529a702863b05a5_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f464fc0f27d5cb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457c862f01de706b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.62, 0.18, 0.18], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54fb03c319d86867_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc4537c692ddaff7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.12, 0.66, 0.16], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7269d72ef9112bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc463a620cf009e5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.29, 0.54, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_119cd2f84a07011d_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a145377de56e4d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.13, 0.44, 0.35, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a46c35fba63812d0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.37, 0.14, 0.3, 0.06, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef939d6075dde301_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a14d0a75b5e6823_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.2, 0.59, 0.09], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66eab9e11cf1e6fe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.2, 0.02, 0.48], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_088451fdb61194d9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.04, 0.72, 0.23], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_465b0032bad586e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_158f4c2ebce804f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b1ca49f3e6c5264_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.15, 0.52, 0.22, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84b34d1b2e577bf2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.7, 0.2, 0.09], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24aaf37feb7d3209_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.2, 0.62, 0.14, 0.0], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57c9825def0e7a2a_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64eb30296601efdd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.15, 0.78, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_511f2ba43686a1e6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.71, 0.1], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eca32a3b923e4da7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.6, 0.39, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b471ea2014e4380e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.12, 0.14, 0.44, 0.18, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39302dd0d428ed0b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.36, 0.06], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac58503902688d22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.19, 0.14, 0.47, 0.09, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab040674edacd800_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.22, 0.64, 0.06], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_267be4a45109324c_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c88659e68e011d37_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66cfe48dc642ea23_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e56fba75e395d19_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff48ee53431e8bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.03, 0.05, 0.14, 0.45, 0.25], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd1ea4d7fb29b2ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c34049788c531d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.15, 0.44, 0.26, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5847daf89d45149_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.6, 0.27, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09fb5273caed0d43_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.08, 0.15, 0.49, 0.18, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6153ab08913c764e_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8cfadf07cbf2a7ab_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.24, 0.09, 0.02, 0.65], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19cc9ce0d60170cb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.42, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a25b3d782f0dd3b_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1f96312d0f0eaf4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.07, 0.3, 0.49, 0.12, 0.01], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccfbdd79ecaf5795_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.36, 0.47, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8d76212c20f55ff_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.27, 0.01], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_704f2150310ff950_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.7, 0.07, 0.15, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f95fe8772f97f41c_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8910c14f60798a2e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.9, 0.04, 0.03], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4f82fbac181400a_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883d64e64503b5e3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.51, 0.41, 0.07], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fd882fdc8921310_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6c3022ddfa9750a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.03, 0.31, 0.2], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd59f193b4a48f96_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5032a3c2370bb2d8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.08, 0.43, 0.39, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_caef6e13019866f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.21, 0.65, 0.08], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8a41bceddc18878_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.11, 0.38, 0.4, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c40d20c8e407380a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.05, 0.14, 0.36, 0.32, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19b9f3224fc3992f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.3, 0.54, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd0f37338c2f6654_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b76e24cb7638271_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.12, 0.38, 0.35, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d5d96be856e6e73_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d472f467babd2c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.1, 0.81, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49cce07ff942abd2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.33, 0.65, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e44f07af9f2bc8c9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.14, 0.65, 0.21], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_520a7417496df3f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.52, 0.06, 0.34, 0.08], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10c19d5ca7961864_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.15, 0.27, 0.42, 0.1, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17c9f3dd0cc7f028_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.39, 0.48, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ca7671b35a6c416_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.25, 0.4, 0.07, 0.21, 0.05, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_598bb2d58a815b87_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.08, 0.41, 0.33], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1814f6cf5d55531_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d26427441ab3e2c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.65, 0.1], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c3c26ceb09d951d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.53, 0.24, 0.22], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ae1e9a779088097_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b99007e3a536ffa9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.15, 0.68, 0.11], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec38a1a0086734c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.5, 0.42, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf07a6e6c4959380_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.08, 0.8, 0.12], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef40381d04b62b78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.08, 0.26, 0.52, 0.07], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_02ddcd5932212adf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.12, 0.43, 0.39, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c42a2dd75c512354_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec6a659e9194cae3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.22, 0.69, 0.05], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11e8b6c8c6ad6012_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.1, 0.48, 0.32, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2f38149cc6f6713_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.46, 0.39, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e222807b809d909_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38693aab8bf70b9a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.27, 0.4, 0.2, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61d6d3e30c30edbe_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_391826797eed8cdc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.31, 0.45, 0.18, 0.01], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3367a439cb75e35a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.52, 0.13, 0.03, 0.32], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faefea2f243d1941_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.59, 0.2], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6adc61752a234abb_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ad920ad1837f01_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.22, 0.46, 0.24, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3cc45bb4c4a3111_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.18, 0.54, 0.21, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db3c7d9c733367c1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.54, 0.22, 0.11], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb35d3e8c0e5c00e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.16, 0.53, 0.16, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab040674edacd800_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dc2429fd1686d79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.24, 0.1, 0.02, 0.64], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4053f1f135fe88cb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.5, 0.34, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_09fb5273caed0d43_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb26c1c3bf097f71_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.0, 0.99], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81039c3b81bad20f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.59, 0.06, 0.3, 0.05], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c169b34fefabe23f_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c189c8ab38de005_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_458f2468a2ebf546_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.85, 0.01, 0.08, 0.06], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a976ef9d46db6951_n", "kind": "noul", "options": ["false", "true"], "target": [0.14, 0.86], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ad07669cfd1dd7e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.31, 0.61, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8bc25af0e15e819_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.8, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9200d8d7b4fd5352_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.3, 0.11, 0.32, 0.09, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_335ac00b28667867_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_818e6a5912f415bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.46, 0.49, 0.01], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ad49bfd98f63290_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.22, 0.59, 0.13, 0.0], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1a262bc94cd1453_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d4be9081a6cb9c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.13, 0.22, 0.44, 0.14, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15cf2fde77f59b21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.28, 0.57, 0.05], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ccc5be05c7b18d1_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08bd7cfe88131a12_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a42d0229c91bd231_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2f455f8f78cafcc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.12, 0.55, 0.24, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c505836d0c959b8_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39302dd0d428ed0b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.29, 0.55, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3d67b8a93aff9f1_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53b6608d177dc514_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0101, 0.2727, 0.6667, 0.0404], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c3ccaf2ab5cdc68_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.49, 0.04], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75304b55a717170a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.18, 0.16, 0.39, 0.13, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_926c675d258b41fb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.72, 0.02, 0.22], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af6878504b38a2de_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.71, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0f93c42cd29312c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.26, 0.1, 0.26, 0.38], "state": "An automation plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8bfa2d9adab1783_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e36ad73d73247b0e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.1313, 0.6869, 0.1111, 0.0707], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dde06d015c82765a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.05, 0.16, 0.47, 0.22, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43305fee0c96062c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.61, 0.35, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_945233792b032c04_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.0, 0.01, 0.92], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf99b9060c7d84ae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.17, 0.16, 0.38, 0.12, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_402b2fed4bd954df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.02, 0.04, 0.26, 0.5, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcc1c231722a9fd7_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c75832a3ac148a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0c127818b7718ed_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.33, 0.14], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d087f591ab7e31df_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8686d3ec4bc7cdb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.26, 0.55, 0.16, 0.01], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c07c8809b00a3dc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.08, 0.44, 0.39, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ffb93e578f0f0313_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.06, 0.17, 0.39, 0.3, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73605d74984c657b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.19, 0.37, 0.26, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccfaf7f9212bdf92_n", "kind": "noul", "options": ["false", "true"], "target": [0.21, 0.79], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f218152faa87316_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bc0432619c91ab6_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6891cb2fb12cb8c3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.67, 0.25, 0.07], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16a7daaf8df5a37c_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85327ca3b3928bfa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.03, 0.0, 0.91], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17bce19bc08b8ea2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.77, 0.16, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdbd27f7f472b992_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.77, 0.07], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db06280c24cbb4bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.23, 0.72, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee340f3175e3ca3a_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f0d8dd4b2a39e29_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d9f1a7e976e1c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.13, 0.59, 0.16, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84bd54b888ba198f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.51, 0.18], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8ce053bf9ae403f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.07, 0.29, 0.5, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_175ae805e63738ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e017d0eb649cb055_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.03, 0.13, 0.64, 0.11], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17a647b6bf2687b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.13, 0.2, 0.38, 0.12, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a10ee444fa8d53c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0ebf81b4c4254a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38f129a707ce886c_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d17d25d4cda37c0d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.07, 0.07, 0.8], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7298ae159b451fac_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94498683df8fe79b_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e332feadb1fcbaa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.48, 0.17, 0.03, 0.32], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26d87b320661dc8f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.82, 0.14, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37eb9a0fe1166b88_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90db561c14637f6a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.16, 0.7, 0.09], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d2a5927935b342_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.42, 0.07, 0.32, 0.06, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b16b6853a83605ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.04, 0.46, 0.47, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fc68c1b88585743_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc5e5ad4dca3ee2c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.05, 0.38, 0.51], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6003a4191b6cd42f_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0b2c820238e7587_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_014932068facba82_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.05, 0.33, 0.53, 0.09, 0.0], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45bc355200ca0d98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.11, 0.69, 0.08], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3ee5dee1bad7bd9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.53, 0.43, 0.01], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c44f4ac71e31ba9d_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35628e238adacf58_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.13, 0.07], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcbb903665e0d8a1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.08, 0.03, 0.86], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90731cb2152555ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.18, 0.45, 0.17, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3ffd7b31c8ae86b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.37, 0.43, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67ec0a7b512f3c9b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.12, 0.85], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd45373dc6ea685f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.2, 0.56, 0.13], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d7723161029d043_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.09, 0.7, 0.16], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b29a490991040093_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.18, 0.15], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a655c1751652bc71_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.25, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8ff7e68636b987f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.32, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88e3fb83c65d665d_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ada55d12ee565707_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.06, 0.23, 0.62, 0.09, 0.0], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f470afbe78692227_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0792432ad7f0fd8_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17d4175415b5aff7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.7, 0.17, 0.11], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22fef1565b8bb896_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56c4064674cedd40_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1293416bf25888ba_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.24, 0.17, 0.4, 0.09, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5e68fd6a2cbdb9f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.34, 0.3, 0.04, 0.32], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6592839acac290a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.47, 0.39, 0.02], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ca5a6d239515c6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.03, 0.94], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4f726433830eb52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.97, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40d9a53d498bf186_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.45, 0.44, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1bcb4199b8cd6f2_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fade8401009d8906_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.48, 0.42, 0.02], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fcecef046e3518b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.65, 0.11], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aece766e405639a_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91feaf377d0a755a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e8764d013d6d050_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.14, 0.59, 0.25, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_181eb395ae4ad552_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.03, 0.22, 0.61, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f960e297488897a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.26, 0.58, 0.06], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec38a1a0086734c3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.35, 0.03], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06558d9fe1864672_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.11, 0.49, 0.32, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec044b3fc949240_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.39, 0.5, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_610e3861aa143625_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.47, 0.44, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_677b6574ee4dfb1a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.43, 0.42, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6da1bbdfa5426797_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.15, 0.44, 0.38, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4143ac2bd2d5ca78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.12, 0.58, 0.23, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97a160edf1344a12_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e084cfa948d2c92_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b71afc446c0a597c_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfec556ccaeaea5e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.11, 0.53, 0.26, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7e55db04f26ca82_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b22b09b9e1ca5ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff2c14058f2a4699_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a90ad2af4fd692e0_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2eb3a3950b3e1b34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.08, 0.45, 0.37, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd7abc4f5bdf73af_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.17, 0.04], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ca8063074990bae_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.16, 0.19, 0.08, 0.57], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f27d6433a3162899_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.52, 0.19, 0.22], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7feba9ca0e8f0abf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.25, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e43971168e23c79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.57, 0.04, 0.3, 0.09], "state": "A proposed workflow parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_53a4bd46613efe83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.2, 0.54, 0.21, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_062db51b690dc390_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.71, 0.06, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_550bec41935f4e32_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e121a7fd620c3bca_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6579746b10a4644_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.35, 0.14], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8f3681f6c1193b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.12, 0.5, 0.31, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_428a40c235dcf7e3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.64, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae5b1a799f0b1779_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.09, 0.16, 0.43, 0.19, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb232c9c0ac2b121_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8db604bb4e97e3c_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_07c6769c5d1dcac7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.23, 0.41, 0.23, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d44bd1077964ed6_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d6087efb0ce8494_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.64, 0.22], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb72a0e06056c95d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.1, 0.53, 0.34, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fd882fdc8921310_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.09, 0.72, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a73a02d5425d9a3_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8d76913dc039378_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.53, 0.45, 0.0], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_757ec56d5379abce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.07, 0.37, 0.5, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_948cb01e11eca4cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.19, 0.5, 0.23, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d4aec78757d71539_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.29, 0.65, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26d87b320661dc8f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.12, 0.57, 0.24, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6647a8a6d9b5af4c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.19, 0.66, 0.08], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97dc77444b285ff0_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72d4871b874a4126_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.31, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ced475407d1bee31_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.53, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6da1bbdfa5426797_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91cf0f67fbb22184_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a1d403c0f5d2ae7_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7484b95f2950c9f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.4, 0.1], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4b3ff45d111c75e_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3dedb18810907edf_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3ffd7b31c8ae86b_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e563cdff265d1b3_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e264baf07527559_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.23, 0.69, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4305e5d1d3c06f01_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8b4a67704e3ad6c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.31, 0.65, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4064e57265f37f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f746f00bb1d89263_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f088784d513cedd2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.13, 0.48, 0.27, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd59f193b4a48f96_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.01, 0.16, 0.82], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf20fe92ac4a9ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0505, 0.0101, 0.0202, 0.1616, 0.6869, 0.0707], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d78e27fc451a3646_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9016528f9181be68_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.25, 0.63, 0.05], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d59ebad8df3fb4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.26, 0.19, 0.37, 0.08, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccedb3e74554e314_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4619a29ab6084ab0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4547e0fc6bd827a_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_863aaf03c6236e8e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.74, 0.21, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59c41fc94cfd5832_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.23, 0.47, 0.21, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_80410555622e2d3d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.13, 0.76, 0.09], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5a2161acb694e75_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e54d9e1bfb2e998_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_899a1381f91590d6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.24, 0.63, 0.05], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca2037718b2c525f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.24, 0.65, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a01dd2aa3738a7e0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.04, 0.16, 0.44, 0.26, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3f08e2e0cc40daf_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64c7d221ad3f7656_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71f93bc0bd3a0c13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.07, 0.32, 0.55, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d88b973563b960f_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec15e95648c8753d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.17, 0.5, 0.23, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44b7ebf8dfb78be7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.28, 0.61, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f759b31ca5c1dec_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.39, 0.52, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fa899a358423b86_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.07, 0.89], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_810ed800d8ef1135_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.34, 0.58, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a16835eef616d23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.13, 0.52, 0.3, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_511c24ed4e9a3d22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.85, 0.15, 0.0], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96ec67e6700d5472_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.29, 0.34, 0.22], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37f72e91ddf9eddc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.41, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7b11e3201c354e5_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6294a7b212ba4e7f_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff20c7792692f2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b957f2f047e50c4a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.42, 0.45, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b4668765d268f40_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.28, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8886147699a1f3ef_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e4ddf6b03025a86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.08, 0.31, 0.5, 0.04], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b13101125ef6e21_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c420e469c6ac521_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.38, 0.19, 0.39], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faf7d96f42026c39_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.51, 0.13], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3944d640adf9ebe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.34, 0.09, 0.01, 0.56], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6366f1c02a22c6a0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.64, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3cd9db5dd8602125_n", "kind": "noul", "options": ["false", "true"], "target": [0.17, 0.83], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc3abd12576d7219_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.08, 0.17, 0.45, 0.18, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e90f92a47b998f6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.22, 0.65, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad13746280d27622_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.16, 0.19, 0.4, 0.14, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61d6d3e30c30edbe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.15, 0.51, 0.26, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41338733d8962b6e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.15, 0.51, 0.24, 0.01], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e152569d4cc0d97b_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f244176d64e3de_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.36, 0.21, 0.02, 0.41], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b879ae8191c0d6d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.89, 0.08, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31dfc80f7da53384_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.53, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f05335e5cb4a6fce_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8eaf138322cf77a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05e965e825eb38e8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.05, 0.77, 0.14], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63c5fb84521bb8a8_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8abeed1e39d4bbb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.41, 0.47, 0.1, 0.0], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e9f440ddadb1bb1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.85, 0.13, 0.02], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c0ac10088c5b0e5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.36, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4963decf96f76b75_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3643edb1f13a73f1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.03, 0.05, 0.17, 0.51, 0.13], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e96b6b16b7999f38_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.43, 0.49, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85cc0503dd474ac1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.33, 0.52, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b558e7427c84d58a_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2a3caa99a4fe6ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5afc55e0c8ecdf55_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.74, 0.2, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a34f40cb83cc6e44_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.33, 0.53, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea6d0c75c9d7f367_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d6491d630e354c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.07, 0.47, 0.41, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1f3182d1df599c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51ab43385f943ace_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.19, 0.35, 0.28, 0.18], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dac15cfc1703e03_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_239bf638e5ec208a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.07, 0.52, 0.37, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23d3ff0e00f463c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c37e33704ad4cac9_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_555c2537120e90ff_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdae166532f73e68_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.65, 0.14, 0.02, 0.19], "state": "An agent plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e73cdd46cc5bafad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.04, 0.5, 0.45, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fa22b6a725a70a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb26c1c3bf097f71_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.4, 0.46, 0.12, 0.0], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7feba9ca0e8f0abf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.05, 0.29, 0.51, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc97878123dd4719_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.66, 0.31, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e42e7fc43ad716c1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.03, 0.07, 0.24, 0.49, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7ae8e9079367878_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4907d3374772f53c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.39, 0.09, 0.29, 0.05, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_085eda9f8c43acf9_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63baae22fbb14cbb_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ccae4b6c9c3c4d3e_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de9e038420e5edbf_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e20f8d372d4969b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.47, 0.52, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d2b96c08abc7ec2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.93, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e38609392eeea46_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.39, 0.52, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, mid-incident. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d70ad19523bd65e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.18, 0.71, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_267be4a45109324c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.12, 0.11, 0.05], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cac519fc57cb0e0d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.2, 0.15], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_deb93ee911710c86_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c40a789196868c88_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.04, 0.66, 0.29], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_545350cc23dba41f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.28, 0.52, 0.07, 0.13], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81e95ae83319c565_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.12, 0.41, 0.24, 0.06], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1859c2a7803f1860_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19bf7912aedb8486_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.16, 0.42, 0.26, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ec27ee2c7c667d3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.37, 0.53, 0.02], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cad86a6e0f307578_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc890de4307f65cd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.66, 0.28, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44e3d31704978752_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.63, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1753f4d3cd5ae888_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.13, 0.19, 0.43, 0.11, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56387005a8233366_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.85, 0.14, 0.0], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f088784d513cedd2_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15dcbcf7841c1b24_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ea4679686ae07d3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.32, 0.02], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18f06aa2a7033b90_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.31, 0.59, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0e31868721b782_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.28, 0.6, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a96d3994cf8175b1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.73, 0.22, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_162c85768348fb98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.12, 0.86], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b4fdee68bf5ccbc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.38, 0.55, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfed58128702df71_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.7, 0.14, 0.15], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b1f4d068952d6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.21, 0.56, 0.05, 0.18], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c80035513f429bbd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.16, 0.2, 0.45, 0.1, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14594dbf022e9748_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.08, 0.37, 0.48, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66cfe48dc642ea23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.22, 0.44, 0.23, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2a39846d1f2a3eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.32, 0.53, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0be9b0ea271c04ea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.11, 0.33, 0.39, 0.08], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45e183baa3a202b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.19, 0.74, 0.06], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a308c45605247bd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.34, 0.53, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_341846233cc4ae0f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.01, 0.12, 0.8], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d44bd1077964ed6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.83, 0.13, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae93774209e1d29f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.05, 0.53, 0.34], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_026cbd92e144aace_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eff659e1a3972b0f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.15, 0.62, 0.18, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e085a5b0dd0e4212_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_107939d9be8dea07_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_168f90213ca83bab_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.43, 0.55, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c278ccf5b704ad23_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.72, 0.17, 0.1], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06c82c93b8b66304_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1f5b7a8156f66d2_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a1528d845079e70_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f853e4d570293112_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.18, 0.21, 0.39, 0.13, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d255094533ec5b16_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.1, 0.43, 0.36], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958bd4d1642f3f2c_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47202d825a764672_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.44, 0.05], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34b520f7f5c47e07_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.43, 0.41, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_900c83af08f9c3df_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.59, 0.03, 0.33, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdbeba76b5f2de50_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.16, 0.45, 0.31, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_deb5e26222a441e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.13, 0.48, 0.28, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e9bf19d0ef72f20_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.22, 0.14, 0.61, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa316cc43372027e_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e67e0c5a2fc3bf6_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f468b1e88e881963_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.07, 0.8, 0.08], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8902342bf666ccc1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.6, 0.19, 0.19], "state": "An automation plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecc3e9af2b753e35_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.35, 0.19], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a97713c1f8d10613_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c26adcfb2d68073_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35f2fcf7a2deb8e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.4, 0.46, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_826e26ecaa12e4ed_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.17, 0.81, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9016528f9181be68_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b29a490991040093_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.2, 0.46, 0.18, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_417c9438787afa31_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.77, 0.13, 0.09], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ddac1f8244e8757_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.41, 0.1], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bb192c0ba816fa9_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86990f33ce63e7a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8c6662a2840fb88_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.05, 0.43, 0.46], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bda017a21524e3db_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fcf880c7bde82819_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.47, 0.44, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef40381d04b62b78_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.42, 0.42, 0.15], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1816c49e6443215a_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b89ad6a04f88023_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4f726433830eb52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.4, 0.43, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a81d184401dc47c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.57, 0.37, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dced8169df28a884_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.94, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8fdbe9c3a61c09a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.15, 0.18, 0.33, 0.19, 0.08], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d86201a38ee571d2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.73, 0.14], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d866c48604f56bae_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.93, 0.02, 0.03, 0.02], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_30b83d5595603d9b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.03, 0.17, 0.6, 0.11], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f58b7d223d8937a_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8e26ce4e14eb363_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.1, 0.41, 0.37, 0.05], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ee4c2a27a1d6ae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.3, 0.55, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f80d66a3be05d890_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.08, 0.49, 0.34, 0.01], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b76d82a44b10201_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.01, 0.96], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bde2f43ac382cbbc_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eda057ce1d539889_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d62d1d6eb099b4b3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.64, 0.04], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41338733d8962b6e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6768, 0.2727, 0.0505], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9ef3836e42bf110_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.29, 0.06], "state": "An agent plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_391826797eed8cdc_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6d55bd8c57867b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.01, 0.11, 0.84], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_423f0dfd3bfc5933_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.03, 0.1, 0.5, 0.29, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcdbd4d1f8d5e174_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.27, 0.11, 0.36, 0.08, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72580fd4b4543676_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32b8f50877956722_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.59, 0.12], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23557d20e60584ca_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.44, 0.48, 0.01], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bab0f0817a36018c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0202, 0.0707, 0.2323, 0.5556, 0.101], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c31fe80e71534ab2_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76d316adfcf10e8f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.17, 0.32, 0.08, 0.33, 0.08, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6fd47ee23167d52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.38, 0.51, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d086f0b9364be664_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.04, 0.48, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10c19d5ca7961864_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.4, 0.08, 0.23], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b027cd418b679b57_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.96, 0.01], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e750dc810a4bf277_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.38, 0.11, 0.28, 0.08, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85d5520848570379_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_779ac12d9306cafa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.39, 0.44, 0.16], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d620400ba389a9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.52, 0.28, 0.18], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81b291c69367ca81_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.24, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea880cef1eb35872_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5045901decfc8a78_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56e452acfcb7b826_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.7, 0.27, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb232c9c0ac2b121_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.61, 0.3, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_172739dc8dc5d65a_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_342ff9f6d8974b86_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.01, 0.1, 0.84], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25f2baf8938fa3f9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.22, 0.44, 0.14, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dea5c5c4509ff59c_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e44f07af9f2bc8c9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.08, 0.5, 0.33], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f087bcd909fb94_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c7934137e6657d08_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc66b9c6410fe1c4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.66, 0.06, 0.17, 0.11], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e47fce249867845_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5ee7b10138ad9b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.21, 0.79], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abb7db2cc34ec3b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.4, 0.57, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_117bddc11d753bc9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.29, 0.54, 0.03], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_826e26ecaa12e4ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9cf791c192f7127d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.54, 0.45, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06a015811e9a4fc2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.46, 0.03], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dba209889718b3ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.25, 0.31, 0.04, 0.4], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6698e6c7d18ea30e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.14, 0.19, 0.35, 0.14, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e18bd2267a4b5075_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.02, 0.18, 0.67, 0.06], "state": "A proposed workflow assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af4aec65a448b8fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_30ede81595d61aac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.15, 0.2, 0.42, 0.11, 0.03], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d274b02ad28389b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.23, 0.36, 0.07, 0.34], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_682327df6e54500a_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f458844b670bcc5f_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da00307b841ddd7f_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_538f93dab892251a_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bae339b4001a7ac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.01, 0.68, 0.25], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0da13b4795671f60_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72576661340036c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.07, 0.23, 0.46, 0.11, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ee2120b0c3593c8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.41, 0.57, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883d64e64503b5e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28e7c55063b9a926_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.1, 0.44, 0.36, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883d64e64503b5e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.11, 0.43, 0.37, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9140f6ac51860c06_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.04, 0.24, 0.6, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_afcc6160291fb4b8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.57, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8351234b700b0ceb_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_631efd049871575d_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6e22164bab248b4e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.16, 0.66, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ee4c2a27a1d6ae_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.62, 0.37, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef8a8463a44fd603_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0404, 0.1313, 0.5455, 0.2626, 0.0101], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c849f0135312bd28_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.33, 0.18, 0.3, 0.06, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6107a1159d65aba_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.57, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd1650b3b3e51360_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.13, 0.47, 0.25, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_271d0ebe45be9b70_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f9767d0d60dbb79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.57, 0.33, 0.01], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67aef4555f5e46bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.27, 0.63, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75405c9a48fe73ed_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.21, 0.22, 0.1, 0.32, 0.11, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ed5f86f56792730_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d2374f68a174632_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efc7522ddc465e6d_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c3c7659d931389b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.74, 0.14, 0.08], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbc2ff581eb4b2e1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.02, 0.1, 0.83], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecf5f18490cd96eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bf36c409e784f4b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.31, 0.55, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_502d8a6158c2f31a_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_739bf8e6e5842277_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.55, 0.43, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bed8d226119c25b_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40a4b720aa7eca0a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.48, 0.32, 0.14, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d1b3d91e66e236d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.33, 0.56, 0.08], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27461670616936af_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.88, 0.06, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec4f9151be116809_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.42, 0.44, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_604c43c018667b44_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.45, 0.39, 0.02], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d998f63507f7b47_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.28, 0.09], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4966af296c2b0d08_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_538f93dab892251a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.29, 0.56, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_604c43c018667b44_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f02eea3042987e21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.51, 0.45, 0.01], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3dcf3fb0292189d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.08, 0.91], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_766aa1893edc931b_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db6e1421cfbc47ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75ae17bbae613c90_n", "kind": "noul", "options": ["false", "true"], "target": [0.35, 0.65], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abf03e6312779830_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.01, 0.04, 0.28, 0.52, 0.05], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_858aa1a2838e1f38_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47ea5da8dfab6b90_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.02, 0.06, 0.43, 0.38, 0.02], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5eadc21fb9e2e6a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.55, 0.03, 0.09, 0.33], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb56c288cd8d34ca_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f81a22a24cd078f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.17, 0.58, 0.19, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9b1a0cabe910127_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.63, 0.1, 0.27, 0.0], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57951842bd33e103_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.34, 0.6, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdbd27f7f472b992_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d33dcbdb17e5180_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.61, 0.3, 0.08], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe57706a36feb7e7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.64, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42e7d70f881666e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7894accfe1731fab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.16, 0.72, 0.07], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e79e0b50df176cae_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c9f7b4481634fa6_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94824439d412b590_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.17, 0.51, 0.12, 0.2], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc8e3a7d7552371e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.29, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d00825e30076c7a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.3, 0.16, 0.31, 0.11, 0.04], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3f244176d64e3de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.13, 0.56, 0.2, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_524ced9c71154d85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.12, 0.57, 0.22, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08dced6b7d058edc_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b77347f4f1490083_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.7, 0.07, 0.01, 0.22], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a193d60eb8372346_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.6, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_402b2fed4bd954df_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b33c9af93a60ee_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f49866d5c9129a42_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.51, 0.07, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4575dd882b0c462c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.56, 0.29, 0.14], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ec701efac4c37b0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.29, 0.61, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95241ca2e8fd5a82_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b4c68ebd28cd2c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A task decomposition uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bf312ca3616442b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.16, 0.19, 0.35, 0.17, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8970eed6a1bafe0f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.46, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10992a7ac04bbf8f_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0ba375513d5c1e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.3, 0.7], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc21f7a35e2e802b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.42, 0.04, 0.01, 0.53], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f15ee87d9c20b25_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3cc45bb4c4a3111_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2581ca94f0fec46f_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_871c63640aae46fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.47, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb19b614dbaa8653_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.23, 0.63, 0.07], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3f41afbbc8f09bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.33, 0.61, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0e23a4df920ebde_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.18, 0.76, 0.06], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c3ccaf2ab5cdc68_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.36, 0.59, 0.03], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c43b254773c18918_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e24576aa6157657c_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_29271f3582cbf5b3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.56, 0.34, 0.01], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fbc3c94fae39a8c_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0dc35d3ebe9f750_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.43, 0.42, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d339320f715d6360_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14e94b5daf0177db_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89626afa95dd213c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.31, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6594a21bf8de74dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.06, 0.4, 0.47, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d7723161029d043_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.73, 0.18, 0.09], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_76e5708240723dd0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.27, 0.48, 0.09], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f550d0c687a97a79_n", "kind": "noul", "options": ["false", "true"], "target": [0.17, 0.83], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1411be854476750_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.13, 0.46, 0.29, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27845f38518fe091_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.04, 0.22, 0.6, 0.05], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed1ba1bf71b18edb_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40532f286c66359f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.18, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f91d1dc39a1eedfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e52b22a99831d74_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.17, 0.4, 0.39, 0.02], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f725e4d9636f3c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33d3dd2a6643c3f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.88, 0.12], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ce3022e17dfde6e_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ae658cb01f7a0de_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1caddd02c110239d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.09, 0.89], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_abb7db2cc34ec3b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.03, 0.16, 0.66, 0.1], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac4425155510fdae_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b960304045286bb_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ad49bfd98f63290_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_699c062e3dad48c7_n", "kind": "noul", "options": ["false", "true"], "target": [0.29, 0.71], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e28a6d2c0ade2843_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.31, 0.11], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a0e33f68fb35503_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.09, 0.33, 0.5, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d315f10ccdde57_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.22, 0.21], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98f655f782787f37_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1531e3b7afb2e3c_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b778c1eb9e0804cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.3, 0.11, 0.36, 0.07, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9f9fa3bc219ec138_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.36, 0.58, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b80bbe3682628153_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.5, 0.3, 0.2], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d498cb39ce5f3e34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.31, 0.56, 0.04], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfc4594c29b7bd64_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e36ad73d73247b0e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.09, 0.28, 0.49, 0.12, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_381f7ed587490a54_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.31, 0.14, 0.3, 0.07, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0fb96a2ce4e9d94_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.41, 0.02, 0.01, 0.56], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e085a5b0dd0e4212_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.18, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab6934e360225844_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7a59cdb0515a9e7_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2145178ee446a7f3_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca94f6743f0238e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.14, 0.5, 0.28, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8ce88e75a3459b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.11, 0.58, 0.23, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d7cd86fc4559a479_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e074c6b049316741_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.13, 0.07, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87e23308981e4f35_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ca7671b35a6c416_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4907d3374772f53c_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da58e7b489dd4554_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b4dacde4eba41b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.09, 0.73, 0.15], "state": "A proposed workflow parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_516eb18cb3238cae_n", "kind": "noul", "options": ["false", "true"], "target": [0.08, 0.92], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_441598af8ded0df4_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5bf992aea4e38392_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dec5aa340555fe3f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.39, 0.52, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71ffb9b7352d17f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.24, 0.13, 0.34, 0.14, 0.05], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7442c0fe0f2a543f_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a08ba0df760fa9f4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.15, 0.17, 0.46, 0.11, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17bce19bc08b8ea2_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a4e159c9fa188fd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.35, 0.33, 0.12, 0.2], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4da478a96ee240d7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.47, 0.14, 0.21], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57a6cfc3de4e31ff_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_894961486eefcb53_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb0a363886e088a8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.27, 0.23, 0.15, 0.35], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958da2319d0cdd38_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.12, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f16d4bac44712e5a_n", "kind": "noul", "options": ["false", "true"], "target": [0.22, 0.78], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10b1b35b103dded2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.03, 0.08, 0.42, 0.33, 0.01], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5507ee782f2e46a6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.92, 0.08, 0.0], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_677b6574ee4dfb1a_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2c6a9998d996b58_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.2, 0.2, 0.36, 0.12, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a86527f4379aaf48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.07, 0.17, 0.45, 0.21, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19bf7912aedb8486_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db452668cf8f9458_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.18, 0.54, 0.23, 0.01], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c61f473f4d7dfa4_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dced71498c2960bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.47, 0.04], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_061c05b4e5348ae0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.91, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d6aa770e70f57ea_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.25, 0.64, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4b3ff45d111c75e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.41, 0.28, 0.29], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a01dd2aa3738a7e0_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b113691808a2ec60_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0104a22dc61ed6a0_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e72dc475ce846af_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb8f6c056dcd1485_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.14, 0.65, 0.12], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c983db6e51a0f2d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.51, 0.28, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24090441f34b8ae1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.37, 0.49, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_feec5a816ad8c4c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.13, 0.2, 0.35, 0.17, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aa82e22dd8820be6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.37, 0.29, 0.28], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5de9d420965a43e5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.09, 0.43, 0.25, 0.23], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e674e3e7840ab7e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.19, 0.67, 0.13], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b71afc446c0a597c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.07, 0.16, 0.49, 0.2, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7140838b6cccc0a9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.64, 0.24, 0.11], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_543a82e57d9c92c5_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7d30904674b2486_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.06, 0.26, 0.54, 0.05], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5757f1b17f3c81_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.14, 0.72, 0.1], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aeba1cb3e4ea934_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.48, 0.11], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63bde75239d2d16f_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c19f0943141555f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.26, 0.07, 0.36, 0.17, 0.07], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6016b048f187618_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.12, 0.57, 0.24, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_341882cab392440a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.54, 0.34, 0.1], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_378109ff13540160_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e86a0f4d6dd519e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.46, 0.5, 0.01], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_353dd2c1fdbcb8fa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.63, 0.24, 0.08, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9fdafc77124012f_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c30825b85491e92_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6536a7a5a135f3ec_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6fa2d67bfee2335_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8fc72382c3f8da4_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1361a0d30b11bdc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.28, 0.61, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a5f478b8cf56baf6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.05, 0.82, 0.11], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_789df7b599d58676_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.11, 0.18, 0.51, 0.13, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6253232b93fd5f52_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.05, 0.87, 0.08], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f6ccca83dcdf267_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.34, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d84cbec2fcb8285f_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fff8d5e8907ea1cd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.79, 0.17, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba8108e3e0d7b1c6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.59, 0.36, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b9a28e56e8d45212_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.61, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a193d60eb8372346_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.79, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93ca9de9da036107_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.54, 0.23, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3d5bdbb2235ec94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.37, 0.53, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34ae804128c31609_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63e1d13d0afdf691_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.43, 0.48, 0.07], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad0a407a4bb2b1c0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.07, 0.87, 0.05], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96ec3a4040407cf5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.06, 0.23, 0.52, 0.15, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5032a3c2370bb2d8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.24, 0.66, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c46ec21becc6e845_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.43, 0.07, 0.03, 0.47], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_342ff9f6d8974b86_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae5b1a799f0b1779_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab4e4a897a07f62e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.18, 0.36, 0.24, 0.05], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec2927ab3f6f40f0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.28, 0.29, 0.04, 0.39], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4abd0003cbf4a7bd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.29, 0.03, 0.57], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_539f5a46052e3b65_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.73, 0.23, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d00825e30076c7a7_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d77c8d902337f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.04, 0.24, 0.58, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61282fd26edc305b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.12, 0.82, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eeff4bf56ae110fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ccc5be05c7b18d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.32, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_497c75b79da47a75_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.27, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb6ff29f04e50920_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.49, 0.36, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c1fa8201759c41b2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.42, 0.44, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a5fe0273aade08a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.23, 0.42, 0.24, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a05f5fdcb436dce_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.27, 0.03, 0.02, 0.68], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20bea058ac26bca9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.06, 0.29, 0.58, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54121102ffa889d4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.56, 0.21, 0.2], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81e9ddf99f2a683a_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An automation plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d92ff1bd865ac6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.32, 0.59, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69210ebcc6ccfdc4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.42, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c9b91d17bed563c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.62, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0465a90fa43140c2_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b414af99f6843843_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.36, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cf3ffe988217525_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.91, 0.06], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66a7767c0bff758_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.33, 0.59, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36356e38be922cb7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.21, 0.44, 0.14, 0.21], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_779dec894044c4a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.2, 0.58, 0.19, 0.01], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff682b592e24c80e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.07, 0.27, 0.51, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e038ff0a6931779_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.33, 0.47, 0.13, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17d3501a65a8d7a2_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_888d3188e4b47d80_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.12, 0.72, 0.08], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d86880153e748db2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0202, 0.3939, 0.5657, 0.0202], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a05f5fdcb436dce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.24, 0.13, 0.4, 0.1, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c17837781fd0663_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.14, 0.34, 0.46, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28eb5ced6398964e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.34, 0.57, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a8e6f8c64e2ff8c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.77, 0.18, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04b32369901a7a1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.14, 0.52, 0.16, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_278e395e631fd4a7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.53, 0.21, 0.06, 0.2], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_864536b5b7330482_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.95, 0.03], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_451ff08e8fd9b65b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.61, 0.18, 0.14], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3e6eed98e0adad8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c47fa69076819b9d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.12, 0.5, 0.29, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9dad05700ff379f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.49, 0.42, 0.01], "state": "A proposed workflow requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ab4cd676c0d9d4a_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e7857b00a4b865a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.01, 0.04, 0.75], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855b343674aa5c6b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.55, 0.3, 0.07, 0.08], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_834929b62c6a1bdf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.59, 0.18, 0.09], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0be14e0f640b501a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.15, 0.52, 0.22, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca3605b67eb614aa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.64, 0.29, 0.06], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6592839acac290a_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_554ddca1aed4c4a6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.28, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96c68a11d667d777_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.31, 0.62, 0.04], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e937637420bb33ff_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.77, 0.11, 0.09], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc0270a376d49151_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd18c81a633c8160_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.24, 0.68, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eda7f78582fe0d1a_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05c83edcd842ad19_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.54, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b2b5132c68f7d83e_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A task decomposition includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33f7e0fb5b177a50_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.82, 0.08, 0.09], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_822f019f3f6ab89f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.24, 0.3, 0.44], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c69086e804f4de1f_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_883c2559626f4e83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.21, 0.27, 0.35, 0.08, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62c978148f357cf8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.47, 0.4, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c46a11442eeaf769_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a95f3d73ec727610_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.24, 0.66, 0.04], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420e4be214e5027f_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64d7d5879157dd9d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.07, 0.14, 0.51, 0.15, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_891b76fb5739a4e1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.06, 0.43, 0.4, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_519819b6824bfe01_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1726f6daf1e8381_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.08, 0.46, 0.37, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fcf46e4e0ae9ba7_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An automation plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1561a885ffa801fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05863c271559edbd_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a1d28c0999ed486_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.38, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1c6731465262a2b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.53, 0.47, 0.0], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_811d577aa5227c54_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.6, 0.18, 0.19], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51a85ba0828b8441_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7630f97850e9cb8e_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2efc4ee6111512b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.55, 0.25, 0.01], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70d6697f81b25fc0_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75816a44b8267847_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.46, 0.43, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6013d43e08494502_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1d15a036295dbc0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.6, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0204c2d3ca675c2e_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An agent plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7fc76ef3c9e8924a_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c96ee0c9f61362b1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.25, 0.65, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e55208dbc6434914_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.61, 0.35, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8851929f7865297a_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b18e67555da367b_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_090ed9fa92f06c50_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.82, 0.01, 0.03, 0.14], "state": "An agent plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db44fb6238b20775_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.11, 0.1, 0.75], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd77b18ae74aa459_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.62, 0.33, 0.0], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d8b29c1ac8e7b08_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_186925c5c7f1b0ad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.22, 0.35, 0.4], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb4aeb00f54597b4_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af1ae3edb8503e9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.27, 0.54, 0.05], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d364f6224d92d0fe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.35, 0.19, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f39069fb8fe8e4c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0202, 0.2727, 0.6768, 0.0202], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_672b283dfebaf19b_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66a4b087937edaa4_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_520da7aa3f73c772_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.01, 0.96], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6087a465590c140b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.69, 0.09, 0.03, 0.19], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277361bd393da474_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c53f00553497363_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c2e042c1af35f52_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.19, 0.16], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c346fd85b3394058_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d7d862a8042f4dd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.69, 0.29, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_175ae805e63738ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.31, 0.6, 0.09], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c9da896b7c32fa9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.63, 0.3, 0.06], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dddaa76826ef1ebc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.21, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4da478a96ee240d7_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8933a7d560cbc9e_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d02f8332e63589a1_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dea5c5c4509ff59c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.05, 0.78, 0.16], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d44f30c48ba2f2fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.53, 0.37, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecbc1c68f3407614_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe6f42ea1e457817_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ea3b35b2df49a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.43, 0.49, 0.01], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cced86e9ea336d82_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7ebac4fc92e459a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.06, 0.33, 0.5, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a67953a7b5447c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.14, 0.52, 0.25, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f14d01e75713386_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.12, 0.38, 0.32, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bc03d253e2fdee5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.23, 0.67, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42eec61568828ed3_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_191412dde51fa9ac_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.54, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_322cc44ab44ce591_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.5, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_900c83af08f9c3df_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69210ebcc6ccfdc4_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44fc15661deebd59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.14, 0.73, 0.09], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51bc82039a5bee02_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.19, 0.05, 0.01, 0.75], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2a39846d1f2a3eb_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_191412dde51fa9ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ee2865018d36731_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.09, 0.47, 0.4, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee564c3af3d6e723_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.08, 0.1, 0.35, 0.29, 0.08], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af4aec65a448b8fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.36, 0.17, 0.28, 0.06, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5391cb894b202033_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.03, 0.54, 0.4], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e968dc3cb4f4db4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.05, 0.35, 0.47, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cce7d2fca842f3cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bfb609893d01230_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a059ba9c436ff76_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.2, 0.64, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_721865a38745e4f4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.76, 0.06, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df38c381467cbed7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.52, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_530d52bca050c182_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.65, 0.11], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7442c0fe0f2a543f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.08, 0.6, 0.19], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_894961486eefcb53_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.38, 0.49, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39080a730e4eb7b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45e183baa3a202b7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.68, 0.02], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fca55cbf1cf8baf_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e6d62420fee6e12_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.63, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d327b5dbbd3805a7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.52, 0.06], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89626afa95dd213c_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533f18762e88cc22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.63, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1754ac0f9a728723_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.05, 0.14, 0.47, 0.27, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f05892fe0a8c7480_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_52852103f9b467c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c3106032e208ef5_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d28771e0a211773d_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f46de790b36fa58b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.55, 0.29, 0.05], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_505ce5360250384e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.13, 0.12, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3d45aa139ff2797_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An automation plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0fc68c1b88585743_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.39, 0.03, 0.04, 0.54], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc2fc02338f02fd6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.16, 0.81, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34bae02ff90ef1e9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.26, 0.72, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ea4679686ae07d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d044bc42a1cba41a_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5c5a7a8115a934c_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c6e43fe666e20628_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.05, 0.13, 0.41, 0.29, 0.04], "state": "An automation plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d620400ba389a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f58b7d223d8937a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.51, 0.21, 0.27], "state": "A task decomposition uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38ea826701498ac6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.17, 0.12, 0.3, 0.41], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ee10e94443f32b2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.85, 0.11], "state": "An automation plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23543d9192e5f9c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.23, 0.25, 0.37, 0.08, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_592823c8399070df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.16, 0.32, 0.17, 0.28, 0.06, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19bfcec860ee3b1b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.13, 0.12, 0.72], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533f18762e88cc22_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8cc9324fc7e00a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.15, 0.46, 0.2, 0.02], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9915302d889050d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.35, 0.6, 0.02], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ae4bb54cf117b33_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b4983998b40db3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.69, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea1ead63b89e9259_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d975f5b3b5d31e59_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.09, 0.44, 0.42, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06a015811e9a4fc2_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e80e032ed2ff03fa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.57, 0.02], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6153ab08913c764e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.08, 0.16, 0.52, 0.18, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f1609e8bc644fe7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.06, 0.36, 0.47, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e56fba75e395d19_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.86, 0.12, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dec5aa340555fe3f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.57, 0.26, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_944f6a27f6834fc7_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_589f7011d258c4f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e554f090acc8542a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.14, 0.83, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f95babb6c57ab22_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.59, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85259c65fde0a38d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.18, 0.53, 0.25, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d99145ed5d7c769_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.32, 0.02, 0.57, 0.09], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_188401eb16308744_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.11, 0.73, 0.12], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_520a7417496df3f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.46, 0.29, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_218e11ba4e8a609d_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e90f92a47b998f6b_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62c978148f357cf8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.39, 0.15, 0.43], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12d381b111292a57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.05, 0.37, 0.52, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_831144c32aca11c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.16, 0.12, 0.36, 0.23, 0.06], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5aab4ab386f4a1a1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.59, 0.02], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e0e31868721b782_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c53d243a75d1d3d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.37, 0.59, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d921cc5a6ba9c4a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.08, 0.41, 0.44, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edb8a8cf5c081e94_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.38, 0.13], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eff659e1a3972b0f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.81, 0.11, 0.08], "state": "An agent plan ends with a human confirmation gate; additionally requests write access it never uses. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4767294664e19455_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c44f4ac71e31ba9d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.52, 0.09, 0.25], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5576a093ebcc7b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bae339b4001a7ac_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c007c734adb75dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7dcaf766b33254f0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.05, 0.93, 0.02], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0e7381c471c3c99_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.06, 0.46, 0.43, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49f3bb6f4fe32cd3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.95, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d84206d71af939_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c24488f1fc4edb1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.41, 0.45, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62db3e8565b45145_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ce42f8ed2f806c_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ae600ace3b488cd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.3, 0.02], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_787952bb1428591c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.47, 0.06], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a151a60a891bd4f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.04, 0.29, 0.56, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8872302717cfdb4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.43, 0.54, 0.01], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_378109ff13540160_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.53, 0.34, 0.13], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c2de955856f4611_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.12, 0.44, 0.34, 0.03], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0932dac066292b13_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12558ba660d576bd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.02, 0.85, 0.05], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82e957cdeee1897a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.36, 0.52, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fede1f7541a5779f_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9af3f58eaa02118_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.01, 0.03, 0.22], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5173494e2c30f87_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.06, 0.65, 0.2], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92d2d03a7eb1d4f8_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c449d87bab1a6f8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.01, 0.18, 0.71, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf4c7bf0d295d29e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.63, 0.23, 0.1], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0a2c38e41e17699_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.08, 0.22, 0.44, 0.15, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2f4ac2c7e18f944f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.01, 0.04, 0.85], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: under a read-only budget, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcb6a245ed646125_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.41, 0.35, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca8dd4cc0c1619dc_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c603c56425afce81_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.16, 0.74, 0.05], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c72fccb90787fa2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.55, 0.33, 0.11], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fa48d936fa4241f1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.3, 0.49, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c58ee4325da4a115_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.49, 0.04, 0.01, 0.46], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ea81dfdb02e9cd9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.56, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2452ff2d94709d86_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.8, 0.04], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb51d311c2f1d68a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.06, 0.32, 0.46, 0.14, 0.01], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f03022ac0ca54301_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.48, 0.24, 0.25], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2b1d867f2d370018_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.3, 0.18, 0.34, 0.09, 0.05], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5dac15cfc1703e03_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.09, 0.6, 0.25, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ee3f41ca2a6305_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.2, 0.14, 0.39, 0.17, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a655c1751652bc71_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.03, 0.09, 0.36, 0.44, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b10fb9fddc3dc254_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.72, 0.23, 0.04], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35782e8cb2da671b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.29, 0.09, 0.33, 0.16, 0.06], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_149225c65f9affd1_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_554dc6f0531144bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fbd5ea0628b7444_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.24, 0.62, 0.08], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eba09f30847e5e78_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.17, 0.25, 0.35, 0.09, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50abb6b0785e2ef1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.53, 0.34, 0.1], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2267478b77d38bff_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.01, 0.21, 0.77], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dd6ef8b42071c0fc_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acb10f5da57d6139_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.15, 0.73, 0.06], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_598bb2d58a815b87_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ba5790ddedb323e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.21, 0.16, 0.37, 0.13, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d7d984399acf0e9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.28, 0.65, 0.04], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6490118a00a2be8_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_783b3030434759c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.26, 0.11, 0.41, 0.09, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_090a6a372ae4df15_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.35, 0.22, 0.33, 0.1], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fe247362fe3b19c_n", "kind": "noul", "options": ["false", "true"], "target": [0.22, 0.78], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54c97a589689f9f1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.06, 0.16, 0.41, 0.27, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da2a643214ced61_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664e57c7080b02a7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.65, 0.17, 0.15], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_544f966967942aee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.16, 0.59, 0.11, 0.14], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24e6ffa6bdce6207_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.2, 0.01, 0.77], "state": "An agent plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db3c7d9c733367c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e5b50b9358a2ac1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.09, 0.43, 0.4, 0.02], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e563cdff265d1b3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.87, 0.09], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6bb37e6e1efa352_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.66, 0.28, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_300909b2ba845114_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1554ec3f6627fb79_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.18, 0.45, 0.25, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a95f3d73ec727610_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.43, 0.08], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c629bc52c4d60e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.17, 0.39, 0.06, 0.3, 0.06, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67ec0a7b512f3c9b_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_428a40c235dcf7e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An automation plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2c5ff337ba296fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab004cf658bb0209_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.08, 0.14, 0.63], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6561aa7f547dce25_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35220e8e83d2c5cd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.32, 0.67, 0.0], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67aef4555f5e46bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b61d59616769564b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.34, 0.48, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d690ab377aab9de0_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a13adf2b2684bb8c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.11, 0.51], "state": "A task decomposition ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f66b13a75b7317df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.08, 0.3, 0.55, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45a25f691bac9246_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.16, 0.26, 0.54], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cebe7973f99e3acb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0101, 0.8182, 0.0909, 0.0808], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3f21c31a92b8684_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.82, 0.15, 0.03], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8671fe913affe906_n", "kind": "noul", "options": ["false", "true"], "target": [0.28, 0.72], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69d438a3a623f30d_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f943ed77e03fb98_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.13, 0.49], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d4e650fbd5c00a8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.43, 0.31, 0.21], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9373c87b21dc0745_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.03, 0.78, 0.06], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e06a74c879aa64_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.08, 0.63, 0.25], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70e2b02436f82194_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c2543e8f1dfc43_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.07, 0.33, 0.44, 0.14, 0.01], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce5f7b177feace42_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48ae61237e67d20f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.41, 0.14, 0.01, 0.44], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ef921b6e17feaeb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0404, 0.0202, 0.0707, 0.2828, 0.5253, 0.0606], "state": "A proposed workflow assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc656df096979d57_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An automation plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_790ac98a21914a6c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.04, 0.18, 0.45, 0.26, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11d6ac10253a0dd9_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e86900e8b47c30d7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.17, 0.34], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb4aeceec6a7dc1d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.46, 0.38, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f35ec43466addbfa_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.08, 0.44, 0.42, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, under reduced staffing. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ddd2ebb8f7f63a2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.03, 0.73, 0.04], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_347861c401fb1ac1_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_271d0ebe45be9b70_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.76, 0.08], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c00a252f79f34100_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d83fe27cb45f5f55_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.42, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9de3c41efd4c166c_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_506a0c9f52174b3f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.01, 0.11, 0.67, 0.19], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc5e620a1d4b512a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.32, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22fef1565b8bb896_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.06, 0.91, 0.0], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b2f47be80766e5c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.46, 0.05], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c49ac7c6a3c733d3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.11, 0.16, 0.73], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9613876fa9a4bb0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.18, 0.17, 0.42, 0.13, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_125d5388a613cc6d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.42, 0.08, 0.25, 0.06, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb69727c2c43faec_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6e8e876991ecf8a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.8, 0.18, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3d5bdbb2235ec94_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.51, 0.36, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6d55bd8c57867b_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86a8c356e378577a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.32, 0.54, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b4fdee68bf5ccbc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.55, 0.43, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_96f77196de8f2658_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb85437c93ff6894_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6de91eb756d4eaee_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.24, 0.66, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0a2c38e41e17699_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea880cef1eb35872_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.27, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c76677ade9a3f58f_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1e32d70bd91fd37_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.39, 0.26], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dcc791ba43cc82d1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.44, 0.11, 0.23, 0.07, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b7eb9f943b36bc6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.36, 0.16, 0.46], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56de72d58106f290_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2ee6eb59258f0f2_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_18000b5f7aa7ec98_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.02, 0.89, 0.06], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c33b0e0b39ed5537_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.31, 0.04, 0.52], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_671ba4dae058fbf9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.31, 0.55, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3f41afbbc8f09bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a hard spend cap, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85e35b49daf9b135_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.33, 0.53, 0.12, 0.0], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d98be830004d47cf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.42, 0.49, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_489e03428fa7a62a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.08, 0.21, 0.4, 0.23, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3c7a78d9e8bb81d_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f0453b14234a47d_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8b98792c440eaa8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.42, 0.41, 0.12], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a7da70dbffd3501_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e24576aa6157657c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.81, 0.13, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_566c7b9721627609_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.09, 0.74, 0.14], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2a82438b5e028cd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.47, 0.11], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d70ad19523bd65e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.41, 0.2], "state": "An automation plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7e55db04f26ca82_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.53, 0.04, 0.01, 0.42], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66a7767c0bff758_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.57, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2b9830e827081b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cf3f424214aca64_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.08, 0.91, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9613876fa9a4bb0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.11, 0.04, 0.67], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_672b283dfebaf19b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.12, 0.72, 0.12], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83544e0b95a7cc8e_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e66820962a5b2331_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.0, 0.02, 0.94], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f943ed77e03fb98_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.22, 0.67, 0.03], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9fcb43c43bc1151_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.16, 0.71, 0.06], "state": "A proposed workflow assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d17d25d4cda37c0d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.08, 0.47, 0.39, 0.01], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5507ee782f2e46a6_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50d5d63319fb5fa5_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42ffe0d486f7774b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.22, 0.54, 0.19, 0.01], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3229945a7f8469e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5ca8198862eba52_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee53640af2d62757_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.43, 0.46, 0.01], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62409e747b7d6f5e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.5, 0.25, 0.21], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43b93adcf8784411_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.41, 0.48, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c5b244facb79004_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1f96312d0f0eaf4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.85, 0.12, 0.02], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f4253883d3e6d55_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.17, 0.46, 0.31, 0.02], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_810ed800d8ef1135_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ef9ea52fd57c913_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14a01012c8c330e3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.32, 0.55, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e2fe20b52051f43_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.06, 0.41, 0.51, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5b71794debfe7dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.51, 0.34, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e109a1030ed525f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.53, 0.46, 0.0], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5d2c673372e2590_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.44, 0.18], "state": "An automation plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de4d647b19942576_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.16, 0.46, 0.32, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9387238920d9c009_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.09, 0.2, 0.42, 0.19, 0.02], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_545350cc23dba41f_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b3944d640adf9ebe_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d6a191ab8d94287_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8671fe913affe906_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.24, 0.68, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f9c40fba04034cf1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.58, 0.33, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66cfe48dc642ea23_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.89, 0.1, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff3359e314da8553_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cd901c762c3bc1b_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf8c654fa856e430_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.41, 0.2, 0.36], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_788197a165507557_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.06, 0.27, 0.52, 0.07], "state": "A task decomposition sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71b8342812e437d6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.16, 0.17, 0.39, 0.18, 0.04], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c8a587b88d26881_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.06, 0.19, 0.47, 0.22, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60dc012f508470e7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.31, 0.48, 0.17, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e78594bfb404a804_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.2, 0.37, 0.31, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a60a14880b05b56_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_933689cf11261639_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56e340df1acc027c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.03, 0.0, 0.96], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f16537bb54ce77e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.61, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95241ca2e8fd5a82_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.02, 0.28, 0.65, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e264baf07527559_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.52, 0.24, 0.07, 0.17], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14754ce6e4f7ece8_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0dc08387db13270c_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08fbec55903b1a53_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9b19509826dc13d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.74, 0.16, 0.07], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_522127d569875724_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "A task decomposition includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_300909b2ba845114_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.38, 0.11, 0.26, 0.08, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfed58128702df71_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.17, 0.54, 0.22, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd1c0a90154413de_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.31, 0.16, 0.31, 0.08, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9c0199a2cd1108d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.2, 0.31, 0.36, 0.05, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f797452272707f42_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.04, 0.91, 0.05], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6ba799d568c4552_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.57, 0.39, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8025f137d326edb7_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b723de0b7be9fc7_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d235e54adbbbb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f9767d0d60dbb79_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25d2386ceb990356_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.67, 0.26, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd20aed82e23d0f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf07a6e6c4959380_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3368ead56aed3e7a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0101, 0.101, 0.404, 0.4646, 0.0101], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bec73f6ad91c788e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.02, 0.87, 0.1], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55a647e60c7d6ae5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.33, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4053f1f135fe88cb_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cdfad2ec70a5c11_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9ef6c4e60169474_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.56, 0.31, 0.01], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b5d5573f3135faa_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4bb5719058fb682_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d245ec7e54735a34_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.57, 0.38, 0.01], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d59ebad8df3fb4f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.28, 0.3, 0.31], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45ab0e3ad2727f59_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a47c0376cf080712_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.07, 0.14, 0.48, 0.2, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3d3a551c157d1cc6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.19, 0.58, 0.12], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c2ff97b6d31a85_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.96, 0.0], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b77cc6e07a66527_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.03, 0.07, 0.23, 0.46, 0.13], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ce2adc22e3a712_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_954ea0f34ed94eda_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.94, 0.04], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4941af120a72dd48_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.23, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2eb3a3950b3e1b34_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bd6c5317d2d45d9_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc890de4307f65cd_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67bbceeaa11de3a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.32, 0.52, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d48370daf7eeb9f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.32, 0.59, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac11a0e4c3654680_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21e6e4c0e6933e23_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6016b048f187618_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.81, 0.17, 0.02], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f117c0f0bf56573a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.07, 0.45, 0.39, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eaf4459ebf6a8843_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdc4890f5e7f634a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.23, 0.3, 0.31, 0.08, 0.02], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26a1f0dfd25aa931_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.41, 0.55, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f3b5418192edda8a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.53, 0.05], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aba648158aa1a6a9_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb2b3a65efad6c0f_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2395188163696e25_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_419562476eb39fd4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.37, 0.42, 0.06, 0.15], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca4a95e0ab35d83c_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e41fc0aee18f1c5a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.49, 0.27, 0.03, 0.21], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c3bbb8f61ba3727a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.2, 0.14, 0.37, 0.17, 0.03], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f102aa5eaa56132b_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_116aa6f2faac581e_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17d4175415b5aff7_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9e08cb9d3dc50f2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.52, 0.13, 0.18, 0.04, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_445550bad035ae08_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.3, 0.23, 0.43], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04badb9a4eb770bc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.45, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69d438a3a623f30d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.46, 0.03], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b904307e7bddbf30_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc76d15a34d3fcc1_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ee2120b0c3593c8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.11, 0.37, 0.43, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6beeb30b00b1f0_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4fbc3c94fae39a8c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.16, 0.08], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1850cf203d2597b9_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a059da170099a97_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d245ec7e54735a34_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.11, 0.07, 0.81], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8e19d2f422c039d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.03, 0.64, 0.3], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ead24e3fa635dd7c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.19, 0.11], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8c505836d0c959b8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.25, 0.58, 0.09], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the quarterly review. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef8a8463a44fd603_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_500e8326fb40eb56_n", "kind": "noul", "options": ["false", "true"], "target": [0.38, 0.62], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c309cbd215074a7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.48, 0.34, 0.14], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd8bc8be1c451998_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.08, 0.3, 0.54, 0.05], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f790be79e1c45d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.57, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6377eb523424c5e8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.41, 0.41, 0.04], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b08100c3159e914_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.26, 0.07, 0.02, 0.65], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca2b45eb62a0d27_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8ee50dff8b05899a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.21, 0.74, 0.04], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b4539bb10a44f15_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74b53dbd2b1be6f1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.83, 0.01, 0.07, 0.09], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1554ec3f6627fb79_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.69, 0.25, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75ba0b4562a1c07c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.17, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7369bfb8e3e4fdfd_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.35, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8b9388062d1208b_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce43627122618cdc_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a06245b6ec7b00_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.72, 0.25, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_50329125c3824eac_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.16, 0.14, 0.67], "state": "A proposed workflow requests write access it never uses; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc5c89de83b26ed5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.43, 0.45, 0.11], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7ebac4fc92e459a_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_06ebc2dbe9754df8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.18, 0.5, 0.25, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c6a684036f1696c_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4954a5186e003a23_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.03, 0.19, 0.57, 0.09], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0a3808fc1abab13_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.09, 0.33, 0.38, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1409082087b1dfd7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.69, 0.02, 0.17, 0.12], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f6bb37e6e1efa352_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c429178549eacfb6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.27, 0.71, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e23ccf0ca89171d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.09, 0.21, 0.52, 0.15, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5268a769c88ab982_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.16, 0.25, 0.05, 0.54], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d02f8332e63589a1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.4, 0.58, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_445550bad035ae08_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61b977e3c75d95f6_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533d3c740b37e234_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dee3ad473e590677_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.18, 0.71, 0.05], "state": "An automation plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533c5b71ebce21b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.03, 0.97], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d975f5b3b5d31e59_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.34, 0.46, 0.11], "state": "A proposed workflow skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a47c0376cf080712_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.37, 0.16, 0.44], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faa9004dd2c31b8a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.6, 0.37, 0.03], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f4053ce0a46581e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.48, 0.2, 0.14], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c44f2f63cddcc8d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.44, 0.23, 0.32], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, following a recent audit. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4c54eee3162a935_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A task decomposition calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c19f0943141555f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.62, 0.13, 0.02, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b5b829bb4db5db4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.6, 0.12], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cedb517c878f3845_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.38, 0.44, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a6e74bd345b9aa2_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b4041529153c34a_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdc4890f5e7f634a_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A task decomposition handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d49d90ec1f220d1a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.24, 0.23, 0.32, 0.09, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_668bc2cafea8dcd6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.1, 0.34, 0.47, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cc36e6bbb370c05a_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_efe61ecf5c494992_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A task decomposition uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e4fc6ddb6691329_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.41, 0.51, 0.01], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4acada9f6b95ebad_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0ec8d6fc5bbc9ce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.11, 0.17, 0.44, 0.16, 0.02], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_622a3b2ec7c5d457_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.37, 0.53, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfe5f723623953f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.07, 0.2, 0.52, 0.18, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2584bf6b9f2e4354_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.04, 0.41, 0.45, 0.1, 0.0], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b414af99f6843843_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.4, 0.46, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a425293af304c507_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6e1f5d4ae235bd6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.14, 0.22, 0.46, 0.12, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_df79e920cf5d75d8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.09, 0.44, 0.37, 0.03], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99190a8116f98039_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f021521c533974d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.14, 0.46, 0.33, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under an active deadline. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d8e390b021d0cad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.17, 0.21, 0.39, 0.11, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1d76834f14ae6d84_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1f7ca004217b52b_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d6087efb0ce8494_n", "kind": "noul", "options": ["false", "true"], "target": [0.86, 0.14], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04ec83834058a5c1_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d886a5259de93355_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.01, 0.02, 0.08, 0.59, 0.22], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b42523d1ff91c900_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.66, 0.3, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0707c19aec893dff_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_548c00fcec078b03_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ce0e6bf6018ad389_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b238f8e8ab1f8f36_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_975191f5cc8e59fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.35, 0.26, 0.38], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_674fa20a6ef65124_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf8c654fa856e430_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bde3a7faf13bca5b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.18, 0.51, 0.25, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aefbdd151872be0a_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aac55707642c142a_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a0f395149b8d83c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.09, 0.49, 0.34, 0.03], "state": "A proposed workflow parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a7ca5a6d239515c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83e6214fba4ab02c_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5be4beaaf631929_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.45, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_480f974ff7a0689a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.64, 0.24, 0.11], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1a262bc94cd1453_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.05, 0.14, 0.48, 0.22, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85d3816d059cec2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d339320f715d6360_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.07, 0.46, 0.38, 0.01], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_591c92803d1f2387_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e48816ba5ec3910_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f170fe50b0271eb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.42, 0.49, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2752683b0ce259f_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bf3e93654c6ae39c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.22, 0.59, 0.12], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e8764d013d6d050_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5578540292ccbcb_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, mid-incident. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6cb2ab5bd90fc36e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.21, 0.29, 0.35, 0.06, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1841f3b777c97023_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_777e864c48b763f7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.45, 0.54, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ded8c953aed34c0_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dced71498c2960bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20e4a2554f969aa6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.27, 0.63, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed1ba1bf71b18edb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.35, 0.35, 0.1, 0.2], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8f7bb9161a36d30_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.46, 0.04, 0.39, 0.11], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75ae17bbae613c90_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.09, 0.35, 0.47, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942db47bbbd7495b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.32, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_777e864c48b763f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.05, 0.3, 0.53, 0.03], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d8e9374eeb0c220_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.03, 0.24, 0.64, 0.03], "state": "An agent plan requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457efabe8101eaa2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.04, 0.91, 0.05], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_61282fd26edc305b_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23543d9192e5f9c4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.16, 0.71, 0.08, 0.05], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3c5da795ad2e5ebc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.74, 0.14, 0.09, 0.03], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef5d004d79878e53_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.28, 0.05], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0143d414f0d3d86a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.39, 0.52, 0.02], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6acb6fa1116ac9f9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.12, 0.5, 0.35, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3bf06088631c40b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.08, 0.65, 0.07, 0.2], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8f7bb9161a36d30_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.16, 0.51, 0.2, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_186925c5c7f1b0ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.56, 0.44], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_755967b0d29f3720_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0, 0.0101, 0.0707, 0.6869, 0.2121], "state": "An automation plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e64cf2928acde53c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.13, 0.49, 0.36, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f138cba703612e1_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3b04691f0b252d89_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7cb9d90d0bcbdf1a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.3, 0.23], "state": "A proposed workflow ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91b615df4bd0f02e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.2, 0.45, 0.23], "state": "An automation plan includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93ca9de9da036107_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.82, 0.13, 0.04], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d946e6895c48574e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.76, 0.09, 0.07, 0.08], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1fac18724759030_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dce96efd497b643c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.46, 0.49, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ec701efac4c37b0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.52, 0.06], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_419562476eb39fd4_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "An agent plan includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e6d62420fee6e12_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_011c67be20284c41_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.41, 0.46, 0.02], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during normal operations. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fca55cbf1cf8baf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.08, 0.77, 0.12], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8fb4fb48c4ae587_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_72576661340036c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56719963804128a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.13, 0.19, 0.37, 0.16, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a533e70855df6cac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.15, 0.47, 0.29, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38ea826701498ac6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.16, 0.5, 0.21, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4848f70f610b5371_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.05, 0.27, 0.58, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e330d12a895ef8a_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An agent plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12d381b111292a57_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd9e3f7b60f05a18_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.73, 0.25, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2a3b43cc3735a19_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.73, 0.11, 0.13, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c995be57ffee36e2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.19, 0.63, 0.17], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17ad7138867e11b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.3, 0.48, 0.06], "state": "A task decomposition handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f4c42ddab15b25c_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57918d94c7408ef4_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_105cf342772e1e93_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.25, 0.66, 0.05], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a913987772d2ede8_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed19ced9352ed0a3_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f30a84c0990011e8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.83, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08e19f869b811d1e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.33, 0.61, 0.02], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b655876bb4420c1b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.15, 0.51, 0.27, 0.01], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5778e094db6dfb27_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.0, 0.02, 0.93], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8fdbe9c3a61c09a9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.46, 0.23, 0.07, 0.24], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c229aed5289b89cf_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.39, 0.05], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54f75a3f9c74fd44_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.11, 0.53, 0.31, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f7ea52256e2c8f6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.07, 0.37, 0.52, 0.03], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71ffb9b7352d17f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8806d4268e4227cb_n", "kind": "noul", "options": ["false", "true"], "target": [0.06, 0.94], "state": "A task decomposition requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_188889171adfd918_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ce80ec26abd5068_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, shortly after a software update. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cccfc9d4693db1fd_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1a5e77e78fe274b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.03, 0.93], "state": "A task decomposition requests write access it never uses; additionally skips the sandbox validation step. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cf4c7bf0d295d29e_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1369ff7780ac5367_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9af3f58eaa02118_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.13, 0.76, 0.09], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a47e0b6a296bccbe_n", "kind": "noul", "options": ["false", "true"], "target": [0.41, 0.59], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ac9e32068f182f5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.1, 0.21, 0.5, 0.12, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6698e6c7d18ea30e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.64, 0.22, 0.12], "state": "An agent plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1bd7ce93aa5572d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e801a98765424a7b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.13, 0.68, 0.06], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_787952bb1428591c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.06, 0.43, 0.42, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d8bf5defe7106c2f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.11, 0.06, 0.82], "state": "A task decomposition calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_990cef6bc329c560_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.16, 0.63, 0.15], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb19b614dbaa8653_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.65, 0.01], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_00d5e20ba0d2e893_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9bc87874aaaa2a25_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.32, 0.21, 0.33], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b25e14c861f3e393_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.42, 0.1], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12c6b11a400dffb3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.41, 0.53, 0.02], "state": "An agent plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98db039c333707d0_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8db604bb4e97e3c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.48, 0.28, 0.18], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2583d2e88741dc32_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bcf267c86ee995aa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.28, 0.06, 0.39, 0.27], "state": "A proposed workflow skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd7dead90f04356a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.19, 0.16, 0.38, 0.13, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bced317f1b3f6edb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.13, 0.16, 0.4, 0.19, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5173494e2c30f87_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.43, 0.51, 0.01], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_16e60230c0b8d60e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.71, 0.18, 0.07, 0.04], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_752aa953c24b7b17_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d327b5dbbd3805a7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0101, 0.2727, 0.6768, 0.0404], "state": "An automation plan requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f16537bb54ce77e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.39, 0.49, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5770a524b2afc506_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.08, 0.45, 0.37, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28596c5a5e2dd012_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.01, 0.02, 0.22, 0.62, 0.04], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f7e3b50d24648695_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.56, 0.4, 0.01], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_405ee0a1961c919c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.17, 0.17, 0.39, 0.15, 0.07], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15420d877f663596_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.06, 0.31, 0.53, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8084730dedaf4734_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d26427441ab3e2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.11, 0.71, 0.17], "state": "An agent plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21000fa3ed8bd88e_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23cd11d993aa249a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.12, 0.14, 0.64], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24573e50cc8e35c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.29, 0.16, 0.35, 0.06, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c42a2dd75c512354_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.8, 0.05, 0.13], "state": "A task decomposition parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cccfc9d4693db1fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.04, 0.1, 0.3, 0.42, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d94f6000e5bfb82_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.06, 0.89, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e152569d4cc0d97b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.33, 0.64, 0.03], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, shortly after a software update. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36aef6185ba7547a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0, 0.0202, 0.1717, 0.6667, 0.1313], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0faaee4227814672_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d61d185c5e2b23d1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.32, 0.2, 0.45], "state": "An automation plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_344d617d1921e28a_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4c525a0debd8ed76_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a8d045508aa2974f_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0fb96a2ce4e9d94_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6933df6e028c48e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.44, 0.52, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d35e59f8edd8f5db_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f26cdee9b43e83a_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fc9620410df9d55_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.53, 0.04, 0.01, 0.42], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e5b2eaa803a5c77_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.71, 0.25, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2348123dcbcfe028_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.38, 0.52, 0.03], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the system migration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_855817f0f262df92_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.04, 0.52, 0.28], "state": "A task decomposition sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: under a read-only budget, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_038509c3e84c2d2f_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b01daeff4baa3d34_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.71, 0.04, 0.24], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_421705ea38b712d0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.21, 0.59, 0.2], "state": "An agent plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d388da3704f721_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.06, 0.33, 0.52, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_226bec6212fdabed_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba3afe1ec466d834_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.38, 0.23, 0.35], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82ae67422f28fa5a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.77, 0.1], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e13cb220fc6916b5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.19, 0.66, 0.08], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b8304f82804214e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.36, 0.22, 0.06, 0.36], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c3c7659d931389b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.06, 0.25, 0.42, 0.23, 0.02], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4767294664e19455_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.2, 0.37, 0.42], "state": "An automation plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_789df7b599d58676_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.71, 0.08, 0.17], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93ebf4b4dfb23aa8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.01, 0.98], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b93387ffcd5495e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0101, 0.0202, 0.202, 0.5657, 0.1919, 0.0101], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b3541100f990e97_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.73, 0.19, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_510ecf4fed3932ed_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1f7ca004217b52b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.25, 0.55, 0.14, 0.01], "state": "A proposed workflow includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c355a3f503bfe39_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4cc2853d8b6012d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4b6fbd9ee92d17b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.43, 0.55, 0.02], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0a2b29cf61d3b1e_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a425293af304c507_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.34, 0.57, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d8b29c1ac8e7b08_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0404, 0.0, 0.0202, 0.9394], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebb7dbc52e7cb1c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.22, 0.51, 0.19, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6d2d05172a1e447_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_54a5968f9c2d26c9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.25, 0.7, 0.03], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31b46eef4935ec18_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.06, 0.44, 0.42, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2ea9d3aa0abde85_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An agent plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_417c9438787afa31_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_353dd2c1fdbcb8fa_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_854b8a74e03195c4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.2, 0.6, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d53c5f1c36e0c9e_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a60a14880b05b56_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.3, 0.32, 0.05, 0.33], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e86900e8b47c30d7_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "An automation plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_707f6746805be952_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e3d915ee67f10c67_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.18, 0.26], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4005eb7b441c70cc_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e576acfd138d738_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.93, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5498145fe6a54907_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.04, 0.85, 0.08], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fc193c3d3bd5b5af_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.19, 0.45, 0.23, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d45502fe97a8ed9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.05, 0.61, 0.32, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a49c31df3fce37a_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during an external partner integration. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ddd2ebb8f7f63a2_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2e8b81ecaae3da5a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.85, 0.09, 0.05], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13799d7fa88ab187_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b6bb19e6ba706f4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.36, 0.11], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d3bf06088631c40b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.24, 0.22, 0.38, 0.05, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12b22ed7fc5446b5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.74, 0.25, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b723de0b7be9fc7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.15, 0.41, 0.37, 0.03], "state": "An agent plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89a61201f80dfb15_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.0, 0.09, 0.75, 0.16], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, mid-incident. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_41d12b2289026310_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.45, 0.43, 0.02], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd20aed82e23d0f4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.31, 0.05], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47524887f080b56c_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5eadc21fb9e2e6a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.34, 0.59, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8fed5a99b63e233_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.34, 0.51, 0.1, 0.0], "state": "A proposed workflow requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f05335e5cb4a6fce_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.24, 0.69, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8dad925a0816bf17_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_838f852d34e0be1a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.41, 0.48, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d52ca6e5e78a8ded_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.04, 0.32, 0.55, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b7eb9f943b36bc6_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ae5b1a799f0b1779_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.29, 0.4, 0.12, 0.19], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with staffing at reduced levels. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20a3331b7a65035c_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_617188c23d5bf8b7_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af1ae3edb8503e9c_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e99f1c0ed9282c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f868d122bbc08b7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.29, 0.67, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da2a643214ced61_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.2, 0.69, 0.05], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_745c0d9918654486_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4e47fce249867845_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.42, 0.32, 0.26], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08fbec55903b1a53_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.57, 0.41, 0.02], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_497c75b79da47a75_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.67, 0.26, 0.0], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b8da37c3cb5eb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6d2d05172a1e447_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.34, 0.02], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the morning shift. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8fd7ad4d05b8572c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3bc981255183ba8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.59, 0.34, 0.01], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a3bee018d02c6d3_n", "kind": "noul", "options": ["false", "true"], "target": [0.07, 0.93], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6453b5e625ebd1d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.32, 0.6, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc97878123dd4719_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during normal operations. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95d25030f2ae9e31_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.06, 0.22, 0.59, 0.07], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31502f6df0447055_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c629bc52c4d60e3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.52, 0.15, 0.29], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0e404304adb4ba26_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.12, 0.46, 0.4, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e135faaab793be7f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.03, 0.22, 0.61, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_850f830e3f465847_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.84, 0.04, 0.02, 0.1], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42a9187416b77eb3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.17, 0.75, 0.06], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e02d61ae92418c1c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.29, 0.0, 0.01, 0.7], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4acada9f6b95ebad_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.42, 0.04, 0.51], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_107939d9be8dea07_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.78, 0.19, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_844c4b85d09c088e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.05, 0.17, 0.54, 0.17], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35c4096348653c21_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.2, 0.43, 0.3, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_951d66c254d57598_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8851929f7865297a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0101, 0.0303, 0.202, 0.6667, 0.0707], "state": "An agent plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d55a0e9ea5d7c4a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.08, 0.43, 0.38, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69d964d0da328e78_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.32, 0.55, 0.12], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da00307b841ddd7f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.28, 0.05, 0.62, 0.05], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba32c0160d76d762_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_99bab976b8a2fd83_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.27, 0.58, 0.06], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e01e2d70105cb419_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.08, 0.63, 0.23], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_423f0dfd3bfc5933_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.22, 0.29, 0.45], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a75c4ef0ae189577_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, mid-incident. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39872950d2be951a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.6, 0.27, 0.1], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1aacb2f3f0fc7028_n", "kind": "noul", "options": ["false", "true"], "target": [0.89, 0.11], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1754ac0f9a728723_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.52, 0.37, 0.1], "state": "A proposed workflow includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed366cb060396997_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92d2d03a7eb1d4f8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.66, 0.14, 0.17], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28b4983998b40db3_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ed04324ea2f1723a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.59, 0.11], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5118a20b3575ac4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.06, 0.5, 0.41, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cc89e34840ee539_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.37, 0.25, 0.08, 0.3], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1557e3813706c2bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.01, 0.05, 0.45, 0.41], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a73ac89663d2aae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.22, 0.17, 0.32, 0.11, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05e81f95f90fa438_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.28, 0.36], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28e7c55063b9a926_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.48, 0.42, 0.07], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd8bd91b2322284d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.25, 0.2, 0.41, 0.05, 0.02], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbc9344f5decd36f_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8e65e15a7ab9c35_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a5a0819e3530752_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_480f974ff7a0689a_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f9cef3e41a6bb6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.25, 0.72, 0.03], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b99e799868e6991_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.17, 0.13, 0.41, 0.19, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_500b156b9f659e7c_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: for a compliance-critical workflow, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7afdf61d85fad916_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.84, 0.1], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98721a84bc59f172_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.61, 0.17, 0.07], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_961a2f4afce49993_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.04, 0.48, 0.18, 0.3], "state": "An agent plan ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1814f6cf5d55531_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.18, 0.71, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bde2f43ac382cbbc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.05, 0.84, 0.11], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under reduced staffing. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c452c5cefe89f758_n", "kind": "noul", "options": ["false", "true"], "target": [0.83, 0.17], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_301b4b8c515c1ecd_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f16d4bac44712e5a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.72, 0.01, 0.12, 0.15], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_56e340df1acc027c_n", "kind": "noul", "options": ["false", "true"], "target": [0.6, 0.4], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2d0d683794af532d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.16, 0.17, 0.38, 0.2, 0.03], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f75e18403c29c782_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.0, 0.97], "state": "A proposed workflow ends with a human confirmation gate; additionally requests write access it never uses. Context: under a read-only budget, during a scheduled maintenance window. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c80d28fe994323fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.17, 0.54, 0.22, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cec3df879b9b1b73_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b7269d72ef9112bf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.06, 0.92], "state": "An automation plan handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb46f9e08879ff17_n", "kind": "noul", "options": ["false", "true"], "target": [0.1, 0.9], "state": "An agent plan requests write access it never uses; additionally ends with a human confirmation gate. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ebc6eb08cc13b72c_n", "kind": "noul", "options": ["false", "true"], "target": [0.34, 0.66], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84d635b2065ec8a4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.26, 0.11, 0.37, 0.15, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b9e19c2d04d1f3d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.57, 0.21, 0.06], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_32f62341fa282fb3_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c9dc5462c1b53d69_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.17, 0.04, 0.68], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da744e7b59d313e1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.61, 0.26], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74b53dbd2b1be6f1_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a59d3fef0d1a6b33_n", "kind": "noul", "options": ["false", "true"], "target": [0.92, 0.08], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b3edef5da93d2b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.81, 0.04, 0.08, 0.07], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91cb03600b1aa61c_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc8c468571f915b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.26, 0.74], "state": "A task decomposition assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d820d82dd055280_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d99145ed5d7c769_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.06, 0.33, 0.52, 0.05], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_951d66c254d57598_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.41, 0.5, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ced475407d1bee31_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.15, 0.69, 0.09], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63559a1ecb211141_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d7e4f38c531b1c6_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, immediately after go-live. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277bf9874476a3aa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.4, 0.13, 0.02, 0.45], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c39b54f693cd17d5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.25, 0.13, 0.4, 0.1, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aac55707642c142a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.15, 0.18, 0.43, 0.13, 0.03], "state": "An automation plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a7b125275824fb2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.81, 0.07, 0.11], "state": "An automation plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_578c75d233dd48c7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.36, 0.61, 0.01], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d07ac07f4b544a4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.86, 0.13, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10bf0fe91ac4b3af_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.0, 0.01, 0.15, 0.73, 0.09], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3a9b93247e7e7295_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.01, 0.06, 0.7, 0.19], "state": "An automation plan calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca2e18e0a25bede0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.44, 0.27, 0.04, 0.25], "state": "An agent plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958da2319d0cdd38_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_107939d9be8dea07_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.01, 0.03, 0.2, 0.58, 0.06], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d3bd4bcb1ecd027_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.09, 0.35, 0.45, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cd23192908e658ee_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.7, 0.27, 0.03], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13799d7fa88ab187_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.18, 0.19, 0.11, 0.52], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a0682e7fa5e948e0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.69, 0.27, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15a2868fb7e6f49d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.12, 0.22, 0.44, 0.14, 0.02], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb26c1c3bf097f71_n", "kind": "noul", "options": ["false", "true"], "target": [0.36, 0.64], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, following a recent audit. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_48a77b7e4ae53d10_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.24, 0.15, 0.3, 0.1, 0.03], "state": "An automation plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_333ef51e959663e6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0404, 0.8182, 0.1313, 0.0101], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6002eeb5233dee24_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.18, 0.43, 0.32, 0.05, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_86cbef0dcae9ef66_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.11, 0.24, 0.43, 0.17, 0.03], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62256a278a065735_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.37, 0.56, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aece766e405639a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.24, 0.44, 0.13, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14450ab872b67bbe_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.41, 0.51, 0.08], "state": "An automation plan assumes an API quota that expired last month; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420b24b4451181b4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.11, 0.03, 0.1, 0.76], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fb76cdd8e743e2d8_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4ca10360bedefab1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.52, 0.12, 0.01, 0.35], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3ebd40a14cfa4e2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.22, 0.66, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eeff4bf56ae110fa_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.15, 0.18, 0.23, 0.44], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e222807b809d909_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.25, 0.16, 0.35, 0.1, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, during the morning shift. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc60bb7eb2af6aff_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.09, 0.24, 0.46, 0.12, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e29f1286a552c7b_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a4427b786583342_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.27, 0.68, 0.02], "state": "A task decomposition skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d2309c64b94e9998_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_eee63d3463464fe7_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7140838b6cccc0a9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.02, 0.12, 0.63, 0.15], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6bc661d3a4234258_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.1, 0.05, 0.18, 0.67], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_739bf8e6e5842277_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.2, 0.71, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_539f5a46052e3b65_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.15, 0.75, 0.08], "state": "An automation plan assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49f3bb6f4fe32cd3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.04, 0.21, 0.55, 0.11], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b0a285e02941a5c4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.8, 0.11, 0.08], "state": "A proposed workflow sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ee1f2666cdb94ace_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.18, 0.3, 0.11, 0.41], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_543a82e57d9c92c5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.98, 0.01], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ef9decb28db7831_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2c5ff337ba296fe_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.76, 0.22, 0.01], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_848e67ded2abc84e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.65, 0.18, 0.14], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_773018795f0555c9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.38, 0.53, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d6709f41072f3a6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.09, 0.58, 0.26, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e1afbe6ef6d34f17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.02, 0.07, 0.38, 0.47, 0.03], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d3b475572a29f22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.15, 0.21, 0.41, 0.1, 0.02], "state": "An automation plan parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c5dfd0c3a1d1267b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.14, 0.74, 0.08], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5a6cea40d3ef1f48_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.09, 0.79, 0.1], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3e54371ca4e877a5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.09, 0.47, 0.37, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42ffe0d486f7774b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.84, 0.15, 0.01], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b6d4346390bbc10_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.35, 0.63, 0.02], "state": "An agent plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ec73a7d425b6710_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.29, 0.62, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_84fb23aa0100ef2b_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66fd83b38be1e340_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.07, 0.21, 0.43, 0.22, 0.02], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66d77e5905e590bc_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d45502fe97a8ed9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.02, 0.97], "state": "An agent plan sequences approvals before the technical check; additionally requests write access it never uses. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17c9f3dd0cc7f028_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.58, 0.29, 0.12], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70e2b02436f82194_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.39, 0.16, 0.42], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a73a02d5425d9a3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.22, 0.49, 0.19, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66eab9e11cf1e6fe_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.18, 0.27, 0.11, 0.29, 0.11, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_122f793ad020a6e4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.23, 0.59, 0.12, 0.01], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08e19f869b811d1e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.67, 0.32, 0.01], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ddaba8510144067_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9117a2db159b7828_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9c34049788c531d2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.49, 0.39, 0.11], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_677b6574ee4dfb1a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.44, 0.31, 0.24], "state": "A proposed workflow sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2cb7aa54e526b1b9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.26, 0.61, 0.05], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17bc3c1b8b5af404_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.4, 0.03, 0.42], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during peak load. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a35b447793ac5a8c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.37, 0.6, 0.03], "state": "An agent plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e7bbd69c2e4aeea_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, immediately after go-live. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3eeffff5a9955317_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.49, 0.49, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c4de750e43ec9cde_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.38, 0.52, 0.1], "state": "A proposed workflow sequences approvals before the technical check; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38d445f75b4e34f7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.04, 0.14, 0.41, 0.31, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_670b1c8d4e765638_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d9f1e23f7cc8ba44_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad7e39855d62418e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.07, 0.46, 0.41, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73605d74984c657b_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "A task decomposition uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25d050326cd9a179_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.12, 0.45, 0.38, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the quarterly review. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d8b4bcaad3a4417_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.14, 0.31, 0.53], "state": "A proposed workflow assumes an API quota that expired last month; additionally requests write access it never uses. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3641d67b48692fff_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.87, 0.02, 0.05, 0.06], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2199ea34c82ddbfc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.12, 0.22, 0.44, 0.15, 0.02], "state": "A task decomposition ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, following a recent audit. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_579e8bb0bb2860d1_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_12579b4ec91b9110_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.33, 0.04], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70d6697f81b25fc0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.26, 0.09, 0.34, 0.08, 0.03], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9368c35590e1c89a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.57, 0.19, 0.04, 0.2], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_edb8a8cf5c081e94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.04, 0.24, 0.63, 0.03], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_71c2ab960a710b1b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.06, 0.24, 0.52, 0.08], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff20c7792692f2b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.15, 0.14, 0.47, 0.11, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb87d018c2278c4b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.54, 0.12, 0.27, 0.07], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff3613b6a58ada98_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.3, 0.55, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28fa653b2eb7b43c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.14, 0.14, 0.37, 0.2, 0.05], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ff0946e7498a5dcc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.23, 0.67, 0.05], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93181c923f545bf0_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally assumes an API quota that expired last month. Context: under a read-only budget, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55a647e60c7d6ae5_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15324ff852a46665_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e187d6ff600f7e52_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.03, 0.1, 0.4, 0.42, 0.03], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_07c6769c5d1dcac7_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ad22aad8c95dd3d4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.32, 0.04], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47202d825a764672_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.25, 0.44, 0.2, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_748d115ea9d44aef_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.03, 0.7, 0.07], "state": "A proposed workflow skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b6c4d58182afc6e_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c979871eafd72afb_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8f7ea52256e2c8f6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.58, 0.35, 0.07], "state": "An agent plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c429178549eacfb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.35, 0.56, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_64c7d221ad3f7656_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.61, 0.37, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88f9c0b0eb9b5f0d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.46, 0.03], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bce8cdb655f6892f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.71, 0.17, 0.1], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_81039c3b81bad20f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.13, 0.48, 0.31, 0.02], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b8df630f1c89defd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.04, 0.36, 0.48, 0.11, 0.0], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d00b08798b0de41_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.77, 0.1, 0.11], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8b98792c440eaa8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.09, 0.49, 0.35, 0.01], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c47fa69076819b9d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.72, 0.11, 0.01, 0.16], "state": "An agent plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0b63953dd4904fa4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.29, 0.15], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bb8f6c056dcd1485_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.36, 0.08], "state": "A task decomposition requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1401d87832234d3b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.1, 0.35, 0.42, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45dc02802258bc86_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.77, 0.18, 0.04], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a10ee444fa8d53c7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.21, 0.1, 0.6], "state": "A proposed workflow handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5945438dde376d91_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.06, 0.44, 0.45, 0.02], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, while on-call rotation changed. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1fb1a52cde089d99_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A proposed workflow requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5cba376fa60c1709_n", "kind": "noul", "options": ["false", "true"], "target": [0.68, 0.32], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe211f1f0dab214f_n", "kind": "noul", "options": ["false", "true"], "target": [0.13, 0.87], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a05f5fdcb436dce_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28eb5ced6398964e_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A task decomposition uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f03022ac0ca54301_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36c1b8f67ec394a6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.15, 0.56, 0.25, 0.04], "state": "An agent plan includes a rollback step for each mutation; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6b2fc6f878b8b9f5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.69, 0.01], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f5d315f10ccdde57_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0707, 0.0404, 0.1111, 0.5657, 0.202, 0.0101], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_977f892687670541_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.19, 0.25, 0.2, 0.3, 0.05, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d5be4beaaf631929_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.08, 0.51, 0.33, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_19bf7912aedb8486_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.56, 0.3, 0.12], "state": "A task decomposition ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during normal operations. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_420e4be214e5027f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.38, 0.59], "state": "An automation plan sequences approvals before the technical check; additionally requests write access it never uses. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_83ab6e296c9f7d49_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.09, 0.17, 0.48, 0.15, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1cdf27e67ab786c3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.2, 0.26, 0.37, 0.1, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ebcd9e69f04cde8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0202, 0.5455, 0.2626, 0.1717], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3641d67b48692fff_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An automation plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59261b9bffaab22d_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_663f119ca52fe746_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.13, 0.44, 0.03, 0.4], "state": "A proposed workflow handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_357f8f38c41b86f0_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.09, 0.87], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7628f9945a417fc8_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1369ff7780ac5367_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.89, 0.06, 0.04], "state": "A task decomposition handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a95602dcc4285f68_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.19, 0.77, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7bb9036c6ffd8030_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "An agent plan calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_680974644b4329c1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.63, 0.13, 0.02, 0.22], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_729cb0865cfc328c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.13, 0.5, 0.35, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under reduced staffing. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_944f6a27f6834fc7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.63, 0.19, 0.15], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_996f730d39e88438_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.02, 0.9, 0.0], "state": "A task decomposition parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e41320686c776c89_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c98aafcc5e4fb11a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.28, 0.72, 0.0], "state": "A task decomposition calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d95fb11bbbfaace_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, under reduced staffing. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f5fc61d42cc823e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.25, 0.01], "state": "An agent plan includes a rollback step for each mutation; additionally requests write access it never uses. Context: under a read-only budget, during the morning shift. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca94f6743f0238e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.3, 0.18], "state": "A proposed workflow includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f5c2644de48b87f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.41, 0.5, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08e08ca1ec82e480_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A proposed workflow skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_090ed9fa92f06c50_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_277212b08de14879_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.02, 0.04, 0.22, 0.54, 0.08], "state": "An agent plan uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5756a2510700b749_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.01, 0.03, 0.23, 0.62, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d255094533ec5b16_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88e3fb83c65d665d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.57, 0.08, 0.02, 0.33], "state": "An automation plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8bb192c0ba816fa9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.0, 0.97], "state": "A task decomposition skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_55ce42f8ed2f806c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.3, 0.66, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: under a read-only budget, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8ff7e68636b987f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.35, 0.55, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82e957cdeee1897a_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the system migration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4aca142b334ba144_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.05, 0.35, 0.53, 0.02], "state": "An automation plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c67ebfeff6c3464d_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.64, 0.08], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a609bc291f7c944_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.02, 0.28, 0.51, 0.18, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during the system migration. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6d831c6f32694712_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.65, 0.03], "state": "A task decomposition requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fcc08a5f1e7e3fb_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01b61491909363ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.13, 0.54, 0.28, 0.01], "state": "An automation plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6fcab09f8d51e2f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.45, 0.15, 0.1, 0.3], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ebf2e11b85fa1b6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.69, 0.19, 0.11], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f6795aefb4a59e8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.37, 0.58, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_91cb03600b1aa61c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.05, 0.45, 0.42, 0.02], "state": "An automation plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f8a76aedf569fbbf_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.01, 0.73, 0.02], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38559f2b11f4e318_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_586530ef2ad24e6b_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.08, 0.86, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, immediately after go-live. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc7b1148da5dd218_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.04, 0.25, 0.62], "state": "An automation plan sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a239a2045ad73b6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.31, 0.12, 0.35, 0.1, 0.02], "state": "An automation plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47aebb6cfa6a43b7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.35, 0.49, 0.04], "state": "An automation plan includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ef9decb28db7831_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.32, 0.66, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_af1ae3edb8503e9c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.24, 0.12, 0.6], "state": "A task decomposition requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1f81a22a24cd078f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.05, 0.39, 0.5], "state": "An agent plan sequences approvals before the technical check; additionally skips the sandbox validation step. Context: under a read-only budget, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_191412dde51fa9ac_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.26, 0.63, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b77347f4f1490083_n", "kind": "noul", "options": ["false", "true"], "target": [0.33, 0.67], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f7e4a8a3ed27a6c_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An agent plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f751ccd9e906a623_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2bc03d253e2fdee5_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44c3a3df716fd00b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.35, 0.57, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f4cfbab243389aef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.17, 0.16, 0.35, 0.17, 0.05], "state": "A task decomposition handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1753f4d3cd5ae888_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b73a2f1b5608ff5d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.34, 0.56, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6ad7cd62b20d0dc4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.01, 0.86, 0.08], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c429178549eacfb6_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A proposed workflow calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1fac18724759030_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.04, 0.07, 0.87], "state": "A proposed workflow requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_854b8a74e03195c4_n", "kind": "noul", "options": ["false", "true"], "target": [0.09, 0.91], "state": "A task decomposition assumes an API quota that expired last month; additionally requests write access it never uses. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7beb6c6172b153ca_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.14, 0.84, 0.0], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f0bdeae7e396903c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.71, 0.19, 0.07], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ecc3e9af2b753e35_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.05, 0.28, 0.6, 0.03], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_47ad3b788abdbe68_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc5c89de83b26ed5_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d0ec8d6fc5bbc9ce_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.24, 0.32, 0.06, 0.38], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: under a read-only budget, shortly after a software update. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2517c7bfde56b254_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.11, 0.71, 0.16], "state": "A task decomposition skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e56d42be7b345ed9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.34, 0.1], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49de0dbf0922af98_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.08, 0.3, 0.42, 0.16, 0.02], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a hard spend cap, during the system migration. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_168f90213ca83bab_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.39, 0.54, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b18885ad84792403_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_480095c0435e9bf2_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "A proposed workflow parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_239bf638e5ec208a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.65, 0.31, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dfacae87d46dcf2f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.21, 0.5, 0.24, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5da943e8d70859ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef0ac2a7489dfa66_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.01, 0.12, 0.84], "state": "An agent plan skips the sandbox validation step; additionally requests write access it never uses. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_790ac98a21914a6c_n", "kind": "noul", "options": ["false", "true"], "target": [0.46, 0.54], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24478f5c28b98017_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.25, 0.66, 0.03], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_745e1edf59cc8b8d_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e51735c0244eae6b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.17, 0.66, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fca385d03bbb8b60_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.11, 0.87, 0.0], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e5b2eaa803a5c77_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.11, 0.51, 0.36, 0.01], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b44c56d2f74d0dd7_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.06, 0.72, 0.12, 0.1], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_82bfddea204734b2_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_27cc1bb79b15da17_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.24, 0.71, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8baa60c91ac64306_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.28, 0.67, 0.05], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4a9e977023f1a4e3_n", "kind": "noul", "options": ["false", "true"], "target": [0.58, 0.42], "state": "An agent plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under an active deadline. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_17c79ada3b6cf0df_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.14, 0.56, 0.26, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4143ac2bd2d5ca78_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.82, 0.16, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_754c9e8b6d9352c3_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_366cac5f50cbb577_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.02, 0.25, 0.66, 0.04], "state": "An automation plan skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d047073db6dacfc7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.28, 0.18, 0.37, 0.07, 0.02], "state": "An agent plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e62d031a17af2be9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.32, 0.48, 0.19], "state": "A task decomposition sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9051a131e1e9333b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.24, 0.53, 0.09], "state": "A task decomposition includes a rollback step for each mutation; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f9230f1146f55ef_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan handles the failure mode explicitly; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40177b253a9b801a_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6f16537bb54ce77e_n", "kind": "noul", "options": ["false", "true"], "target": [0.27, 0.73], "state": "A task decomposition skips the sandbox validation step; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f844dbc3827c0889_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.33, 0.54, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: under a read-only budget, as part of a compliance sweep. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_006adc6f46720911_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.02, 0.13, 0.75, 0.08], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the system migration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a4e004f033fb6d15_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.14, 0.03, 0.83, 0.0], "state": "An agent plan parallelizes steps that share a lock; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, mid-incident. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ebf2e11b85fa1b6_n", "kind": "noul", "options": ["false", "true"], "target": [0.25, 0.75], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05219f41b5403ab2_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.57, 0.17, 0.21], "state": "A proposed workflow includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca8dd4cc0c1619dc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.01, 0.23, 0.71], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b9cbfbb02b29a85_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.13, 0.19, 0.44, 0.16, 0.03], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6003a4191b6cd42f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.01, 0.06, 0.39, 0.45, 0.02], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec1f535871b5939d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, under an active deadline. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2df20a4352a46d2c_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8700d591eb5992f7_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fec57df729f93aa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.74, 0.19, 0.02], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_05971426306b782d_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "A task decomposition sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c2543e8f1dfc43_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2102be411c2f55b_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c07da8885bf89fb_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.27, 0.63, 0.04], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b993ff2a040bfc64_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally assumes an API quota that expired last month. Context: under a read-only budget, under an active deadline. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fbe2cb93d6ec0290_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3fc0957fb28cd34f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.02, 0.5, 0.36, 0.12], "state": "A task decomposition includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ca65d927dec0d85f_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_813e5c81dc7d40f6_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.33, 0.52, 0.13], "state": "A task decomposition uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7123dd1e6dd40212_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.96, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bc44430dd91c9c6a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.23, 0.47, 0.24, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_13a218bc8aa5a293_n", "kind": "noul", "options": ["false", "true"], "target": [0.62, 0.38], "state": "An automation plan calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3722ed95874190a8_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.27, 0.19, 0.34, 0.08, 0.03], "state": "A task decomposition includes a rollback step for each mutation; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during the quarterly review. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba75ca2ac4dcba72_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.14, 0.76, 0.08], "state": "A task decomposition skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c57f86afbe9b0dd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.15, 0.4, 0.4, 0.02], "state": "A proposed workflow handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_34bae02ff90ef1e9_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1c2e640271afe601_n", "kind": "noul", "options": ["false", "true"], "target": [0.97, 0.03], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: under a read-only budget, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33afe2d1a622d26d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.54, 0.03, 0.27, 0.16], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2ca8063074990bae_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "A proposed workflow ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6e2936559b2bcd5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.1, 0.61, 0.25], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0511140a3ffd75ec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.15, 0.11, 0.66], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_699fa236d4906e01_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9ab0329b0f73a34_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e8a687bbb754f25c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.04, 0.3, 0.56, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, under reduced staffing. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1293416bf25888ba_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_21159ca42479a861_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.06, 0.39, 0.49, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da087aadb2701e6e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.37, 0.53, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a 40-minute deadline, shortly after a software update. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4663b6ae4b1eedc5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.09, 0.42, 0.38, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the morning shift. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_70338ec08652708e_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.67, 0.19, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0aa3991616910d45_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, shortly after a software update. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14754ce6e4f7ece8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.12, 0.19, 0.17, 0.52], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6a0a9c4b94bf5034_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.73, 0.13, 0.1, 0.04], "state": "A proposed workflow includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66fd83b38be1e340_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.73, 0.24, 0.02], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2f455f8f78cafcc_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec2c84877943e2d_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75e882b5549a9e65_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.62, 0.22, 0.16], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4b3541100f990e97_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: with a hard spend cap, mid-incident. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a4e159c9fa188fd_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dce96efd497b643c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.66, 0.33, 0.01], "state": "A proposed workflow requests write access it never uses; additionally sequences approvals before the technical check. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c2fbf3499ed1dc4b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.22, 0.67, 0.09], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during normal operations. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e8131ec88b99f99_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.38, 0.33, 0.26], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, while legacy systems are still active. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e333d2f212cf3936_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.78, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally assumes an API quota that expired last month. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dea5c5c4509ff59c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.26, 0.54, 0.17], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7058def01b14c20_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.15, 0.13, 0.41, 0.16, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a6fc617bfc967933_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.1, 0.31, 0.19, 0.34, 0.05, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: under a read-only budget, following a recent audit. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a696407522a9bc64_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1855d2e3a8897bf3_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.71, 0.27, 0.01], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, with staffing at reduced levels. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f1b0cc201477bd56_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.15, 0.12, 0.38, 0.23, 0.06], "state": "An agent plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_37a423670bdccac0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.34, 0.11, 0.36, 0.06, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b6490118a00a2be8_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.04, 0.18, 0.7], "state": "An agent plan parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4428f4cb876dda16_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "A task decomposition requests write access it never uses; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the system migration. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_08b9fbdc44f3716e_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "A task decomposition requests write access it never uses; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ac9e32068f182f5_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An automation plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49a02484f2ff2471_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.5, 0.44, 0.02], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_341846233cc4ae0f_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6c5d1498f6c4a4c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.14, 0.13, 0.43, 0.16, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, with staffing at reduced levels. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac7ee7a976c48eef_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_879c368a61ddf125_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd70b14f59d81f96_n", "kind": "noul", "options": ["false", "true"], "target": [0.52, 0.48], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a439926d35bddc7b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.39, 0.53, 0.07], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bfefec13b8322d6d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.04, 0.28, 0.6, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2e06a74c879aa64_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.37, 0.12, 0.27, 0.07, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8e1018185e1de50e_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "A task decomposition calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90b951b396fe70c9_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "An automation plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44c2397332e5ce85_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.0, 0.0, 0.97], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6324d0694e9e255_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.2, 0.32, 0.11, 0.3, 0.06, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fa9730d25309352_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ebf2e11b85fa1b6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.14, 0.5, 0.3, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_024ace9bc5ba2e79_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.12, 0.86], "state": "An agent plan requests write access it never uses; additionally sequences approvals before the technical check. Context: with a hard spend cap, during normal operations. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_44b7ebf8dfb78be7_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.03, 0.4, 0.52, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_873b5e2860432e39_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6fc5913112177801_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_15a2868fb7e6f49d_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, shortly after a software update. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2584bf6b9f2e4354_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.74, 0.22, 0.03], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, at the end of the fiscal period. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c16bc831c9fd5cd3_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.05, 0.17, 0.68, 0.08], "state": "An agent plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during peak load. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e9347e96726f4d06_n", "kind": "noul", "options": ["false", "true"], "target": [0.77, 0.23], "state": "A proposed workflow sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0baca6dfb1c8ed56_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43ee37918ac34570_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.64, 0.29, 0.07], "state": "An agent plan calls the refund API before checking dispute status; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c92e0c387f352af_n", "kind": "noul", "options": ["false", "true"], "target": [0.63, 0.37], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cb29893d55dccaeb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "An agent plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, with a new vendor onboarded. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_446546c30accb6bc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.07, 0.2, 0.42, 0.25, 0.02], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the system migration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c34d33694f626a94_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a08ba0df760fa9f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.48, 0.52], "state": "An agent plan parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b9bc07380cc09b4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.13, 0.51, 0.26, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f62571168dd90923_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during an external partner integration. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36aef6185ba7547a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.52, 0.29, 0.17, 0.02], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_228949039af8c595_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e86a0f4d6dd519e4_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during normal operations. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_457c862f01de706b_n", "kind": "noul", "options": ["false", "true"], "target": [0.49, 0.51], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: under a read-only budget, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e7504ae3bf5d8443_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.06, 0.14, 0.42, 0.28, 0.03], "state": "A task decomposition sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec317dbe447ecba1_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.09, 0.04, 0.71, 0.16], "state": "A task decomposition ends with a human confirmation gate; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6ff3adf6afd9932_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.05, 0.38, 0.44, 0.11, 0.01], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: under a read-only budget, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42c3eaf0ad0be25f_n", "kind": "noul", "options": ["false", "true"], "target": [0.59, 0.41], "state": "An agent plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, mid-incident. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7e113e6827c63974_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.14, 0.43, 0.04, 0.28, 0.07, 0.04], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_db6e1421cfbc47ee_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.11, 0.55, 0.29], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, immediately after go-live. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1d235e54adbbbb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.0, 0.02, 0.43, 0.54], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, immediately after go-live. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9e206d508c217e03_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.3, 0.64, 0.03], "state": "An agent plan uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: with a hard spend cap, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f42a9965997ac00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.05, 0.15, 0.45, 0.3, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, following a recent audit. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_670b1c8d4e765638_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.13, 0.14, 0.2, 0.36, 0.15, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d6e350c427ae30a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.34, 0.53, 0.13], "state": "An automation plan ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, immediately after go-live. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_550bec41935f4e32_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.05, 0.12, 0.49, 0.26, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_57d6b667a8d0e2e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.25, 0.17, 0.31, 0.13, 0.03], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b5600d2a6da2baba_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.97, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f2dd4ebb8f240660_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: with a hard spend cap, while on-call rotation changed. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9da8db47b84fe841_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.56, 0.43, 0.01], "state": "A task decomposition assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94e48928ed68a41c_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "An automation plan assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_39c23f5624930a35_n", "kind": "noul", "options": ["false", "true"], "target": [0.4, 0.6], "state": "A task decomposition includes a rollback step for each mutation; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e33940d3dbb79e0b_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.38, 0.6, 0.01], "state": "An agent plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea021e551d8aa539_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.2, 0.73, 0.07], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8309b3145c160337_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.0, 0.11, 0.88], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ddbb4775aff4ab3_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d95067d6d81c028_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.36, 0.58, 0.06], "state": "A proposed workflow skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ec4f9151be116809_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b124407b91a3a718_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.03, 0.08, 0.45, 0.36, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during an external partner integration. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b45a658e0ec00ad1_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c8bdde13c06e1163_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.19, 0.47, 0.21, 0.01], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8fd7ad4d05b8572c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.27, 0.73, 0.0], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1ed5d894a365ff88_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.03, 0.36, 0.19, 0.42], "state": "An agent plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de146d67ec369d6c_n", "kind": "noul", "options": ["false", "true"], "target": [0.51, 0.49], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ca8a692922bb9ec_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.09, 0.91, 0.0], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_415b32b9c52d106f_n", "kind": "noul", "options": ["false", "true"], "target": [0.64, 0.36], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during peak load. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5c30825b85491e92_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.12, 0.32, 0.54], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_25cafd0a29d9d807_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.59, 0.11], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6c4c29b0ba9cd2a4_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.06, 0.17, 0.61, 0.16], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_576083f6f0f02f51_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.48, 0.16, 0.35], "state": "A task decomposition assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: with a hard spend cap, mid-incident. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3ff48ee53431e8bf_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3cd9db5dd8602125_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.5, 0.02], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0f26cdee9b43e83a_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.54, 0.07], "state": "An automation plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9ec044b3fc949240_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6e1f5d4ae235bd6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.67, 0.11, 0.11], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1794d9cbd14d2f7d_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.24, 0.37, 0.13, 0.26], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_66a4b087937edaa4_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.0, 0.01, 0.06, 0.69, 0.21], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_682327df6e54500a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.15, 0.3, 0.34, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, immediately after go-live. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d6324d0694e9e255_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_97b1a72f226f8add_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.01, 0.02, 0.21, 0.6, 0.07], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, with a new vendor onboarded. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43e79444e0d58405_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A proposed workflow calls the refund API before checking dispute status; additionally parallelizes steps that share a lock. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0ee7760733d7a5e6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.03, 0.53, 0.43, 0.01], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e684df25821a7ae5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.24, 0.67, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9fcc08a5f1e7e3fb_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.06, 0.92], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_51bc82039a5bee02_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.21, 0.28, 0.11, 0.34, 0.05, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c061b9c3430adba3_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.35, 0.64, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during normal operations. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_93967484cf776134_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.12, 0.42, 0.42, 0.01], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da08c550f2eb0000_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.04, 0.64, 0.22, 0.1], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e4f8a800dc49aca1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.02, 0.03, 0.13, 0.63, 0.12], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, following a recent audit. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6013d43e08494502_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.3, 0.43, 0.19, 0.08], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally parallelizes steps that share a lock. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bbe20ef5ebe63435_n", "kind": "noul", "options": ["false", "true"], "target": [0.94, 0.06], "state": "A proposed workflow includes a rollback step for each mutation; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_942db47bbbd7495b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.12, 0.47, 0.39, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_186925c5c7f1b0ad_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.39, 0.52, 0.03], "state": "An automation plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: with a hard spend cap, during the quarterly review. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_994f64a382cf1268_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.16, 0.45, 0.37, 0.01], "state": "A proposed workflow calls the refund API before checking dispute status; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94faafcb35e7a3b4_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.53, 0.37, 0.09], "state": "An agent plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ef90bf52e81314da_n", "kind": "noul", "options": ["false", "true"], "target": [0.84, 0.16], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0100824433beb91a_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aaec8b5eaa79ad4f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.05, 0.93], "state": "An agent plan requests write access it never uses; additionally handles the failure mode explicitly. Context: with a hard spend cap, during peak load. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7a543b2067975acc_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.35, 0.55, 0.05], "state": "An agent plan calls the refund API before checking dispute status; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7f0afdd191626b26_n", "kind": "noul", "options": ["false", "true"], "target": [0.76, 0.24], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during the morning shift. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c420e469c6ac521_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.11, 0.23, 0.23, 0.36, 0.06, 0.01], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69d964d0da328e78_n", "kind": "noul", "options": ["false", "true"], "target": [0.44, 0.56], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_75304b55a717170a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.2, 0.27, 0.05, 0.48], "state": "An automation plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90fb8ac3fa895a3f_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: with a 40-minute deadline, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c00c0f6e22d6e00f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.27, 0.71, 0.02], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_797ad98a05e1e0c9_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.43, 0.44, 0.01], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_709d241c6cc26a01_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.63, 0.28, 0.09], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, under reduced staffing. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a3d5c8a702c24596_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.21, 0.72, 0.05], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, during the morning shift. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_745e1edf59cc8b8d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.13, 0.31, 0.49, 0.05], "state": "An automation plan handles the failure mode explicitly; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the system migration. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c251f31c548527da_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.01, 0.02, 0.16, 0.73, 0.04], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a104cb0d14b062fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.86, 0.14, 0.0], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ba20009e5f08de00_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.04, 0.1, 0.52, 0.31, 0.01], "state": "A task decomposition estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_59261b9bffaab22d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.02, 0.08, 0.52, 0.35, 0.01], "state": "A proposed workflow sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_92020e4efe4f1118_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.39, 0.6, 0.01], "state": "An agent plan requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during a scheduled maintenance window. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_047f4022d0ade92b_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "An agent plan calls the refund API before checking dispute status; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4f2da9288a69331e_n", "kind": "noul", "options": ["false", "true"], "target": [0.55, 0.45], "state": "An automation plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during an external partner integration. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d1cb3dfc7166aa6_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.46, 0.42, 0.12], "state": "A task decomposition assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_664d0708dabb0a71_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.03, 0.55, 0.38, 0.01], "state": "A proposed workflow requests write access it never uses; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b686d257c728d7ae_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "An agent plan estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_67c7fd74dbcff67e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.33, 0.56, 0.03], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8199c5974489484e_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cad86a6e0f307578_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.01, 0.04, 0.37, 0.52], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: with a 40-minute deadline, mid-incident. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4d5daf40975f0e64_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.02, 0.87, 0.09], "state": "An automation plan handles the failure mode explicitly; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, under reduced staffing. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac7ee7a976c48eef_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.05, 0.15, 0.33, 0.36, 0.03], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b2039adbd28ea39_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "A task decomposition sequences approvals before the technical check; additionally skips the sandbox validation step. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7167cb4771b91b94_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.03, 0.2, 0.71, 0.04], "state": "A proposed workflow ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during peak load. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7af3356850512d7c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.07, 0.45, 0.36, 0.12], "state": "A proposed workflow includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: under a read-only budget, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3313e203b53cdcfe_n", "kind": "noul", "options": ["false", "true"], "target": [0.54, 0.46], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3367a439cb75e35a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.04, 0.07, 0.35, 0.41, 0.07], "state": "An automation plan handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_45bc355200ca0d98_n", "kind": "noul", "options": ["false", "true"], "target": [0.66, 0.34], "state": "A proposed workflow skips the sandbox validation step; additionally ends with a human confirmation gate. Context: under a read-only budget, during peak load. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a2b9830e827081b9_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.75, 0.14, 0.1], "state": "An automation plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: under a read-only budget, during the quarterly review. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_433958e5af47da38_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.74, 0.22, 0.04], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, while on-call rotation changed. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6dc392d28da02276_n", "kind": "noul", "options": ["false", "true"], "target": [0.19, 0.81], "state": "An agent plan sequences approvals before the technical check; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during the holiday freeze. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d044bc42a1cba41a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.19, 0.2, 0.37, 0.1, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8b9e19c2d04d1f3d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.17, 0.19, 0.14, 0.39, 0.09, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: with a hard spend cap, with a new vendor onboarded. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_04b32369901a7a1e_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.14, 0.35, 0.04, 0.47], "state": "A proposed workflow parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6da1bbdfa5426797_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.49, 0.41, 0.1], "state": "A proposed workflow assumes an API quota that expired last month; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, following a recent audit. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_95d25030f2ae9e31_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "A task decomposition ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, immediately after go-live. Reported by the supervisor.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5f6ccca83dcdf267_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.02, 0.44, 0.52, 0.01], "state": "An agent plan assumes an API quota that expired last month; additionally skips the sandbox validation step. Context: under a read-only budget, while legacy systems are still active. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8001eab2d141ee25_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.08, 0.24, 0.57, 0.09], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0d33dcbdb17e5180_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan sequences approvals before the technical check; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3aca263e8d939555_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.06, 0.16, 0.42, 0.25, 0.03], "state": "A proposed workflow sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: under a read-only budget, immediately after go-live. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_74a03ad9873cae51_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.78, 0.15, 0.06], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_4575dd882b0c462c_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0faaee4227814672_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.05, 0.13, 0.34, 0.4, 0.04], "state": "A task decomposition requests write access it never uses; additionally includes a rollback step for each mutation. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_89c00a58a35cc70e_n", "kind": "noul", "options": ["false", "true"], "target": [0.39, 0.61], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally skips the sandbox validation step. Context: under a read-only budget, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2c8cc2539136eeb6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.12, 0.3, 0.12, 0.38, 0.07, 0.01], "state": "A proposed workflow handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9186f39e644d0c75_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.46, 0.28, 0.1, 0.16], "state": "An agent plan handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the quarterly review. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a1cd9e71d4018ede_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.03, 0.17, 0.72, 0.07], "state": "An automation plan parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cccf80115099d7be_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.11, 0.18, 0.41, 0.19, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, during normal operations. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_26ce5db2672f4f7c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.99, 0.0], "state": "An automation plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a9ebfaed888061d0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.47, 0.44, 0.09], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally calls the refund API before checking dispute status. Context: with a hard spend cap, during the quarterly review. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_78b61e44ca22dd07_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.85, 0.09, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, immediately after go-live. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d1e62540d9d699f5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.05, 0.24, 0.02, 0.69], "state": "An automation plan parallelizes steps that share a lock; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_cfa876310e2169d5_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition skips the sandbox validation step; additionally handles the failure mode explicitly. Context: under a read-only budget, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85d3816d059cec2f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.44, 0.05], "state": "A task decomposition assumes an API quota that expired last month; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during peak load. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd0d86ebb49f4451_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.13, 0.09, 0.32, 0.46], "state": "An agent plan ends with a human confirmation gate; additionally includes a rollback step for each mutation. Context: under a read-only budget, while on-call rotation changed. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_708b0c3ae85e9cbc_n", "kind": "noul", "options": ["false", "true"], "target": [0.69, 0.31], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally handles the failure mode explicitly. Context: under a read-only budget, during the quarterly review. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87071155e73162a4_n", "kind": "noul", "options": ["false", "true"], "target": [0.93, 0.07], "state": "An automation plan skips the sandbox validation step; additionally requests write access it never uses. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fcf880c7bde82819_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.52, 0.44, 0.04], "state": "A proposed workflow assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_23d3ff0e00f463c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.07, 0.56, 0.34, 0.01], "state": "A task decomposition requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_22a69990af1c557f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.36, 0.01, 0.62, 0.01], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, during peak load. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5d4be9081a6cb9c2_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.16, 0.06, 0.31, 0.47], "state": "A proposed workflow ends with a human confirmation gate; additionally sequences approvals before the technical check. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0465a90fa43140c2_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.29, 0.59, 0.04], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, under an active deadline. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9857bad442b7f52a_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.68, 0.28, 0.02], "state": "A proposed workflow calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7c3f04724692fb0b_n", "kind": "noul", "options": ["false", "true"], "target": [0.82, 0.18], "state": "A task decomposition skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_332242cd51ed0675_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.01, 0.96, 0.03], "state": "An automation plan sequences approvals before the technical check; additionally estimates 2 hours for a task with a known 2-day dependency. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the automated check.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_98b1ea440d746490_n", "kind": "noul", "options": ["false", "true"], "target": [0.21, 0.79], "state": "An automation plan calls the refund API before checking dispute status; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3643edb1f13a73f1_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during the system migration. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2fc04f26d489b0c1_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.51, 0.46, 0.03], "state": "A proposed workflow calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, during a scheduled maintenance window. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_362a4b974ba77252_n", "kind": "noul", "options": ["false", "true"], "target": [0.85, 0.15], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10c02a9cb53dfaaa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.54, 0.42, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b75b8cdbdc3be5f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.04, 0.45, 0.45, 0.02], "state": "A proposed workflow requests write access it never uses; additionally calls the refund API before checking dispute status. Context: under a read-only budget, during an external partner integration. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7308bf3f0058c13c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.55, 0.31, 0.13], "state": "An automation plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during normal operations. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_be6bb4e90c01ef57_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.44, 0.16, 0.05, 0.35], "state": "A task decomposition handles the failure mode explicitly; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, under an active deadline. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f621483a700a1399_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0202, 0.0505, 0.202, 0.5657, 0.1515, 0.0101], "state": "A proposed workflow ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_36b736b87b466f18_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a hard spend cap, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_24cf09a2c39c5017_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.06, 0.02, 0.04, 0.18, 0.63, 0.07], "state": "A task decomposition parallelizes steps that share a lock; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the holiday freeze. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1e4c3d3e24670484_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.11, 0.42, 0.38, 0.07, 0.01], "state": "An automation plan ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8a84e85eb988ff68_n", "kind": "noul", "options": ["false", "true"], "target": [0.31, 0.69], "state": "An agent plan ends with a human confirmation gate; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_faa9004dd2c31b8a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.02, 0.06, 0.38, 0.47, 0.02], "state": "An agent plan requests write access it never uses; additionally parallelizes steps that share a lock. Context: with a hard spend cap, during the holiday freeze. Reported by the monitoring system.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_791c3e52311209ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.37, 0.63], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea2ead956484abce_n", "kind": "noul", "options": ["false", "true"], "target": [0.04, 0.96], "state": "A proposed workflow requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: with a 40-minute deadline, mid-incident. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fd4c2d46092907cf_n", "kind": "noul", "options": ["false", "true"], "target": [0.53, 0.47], "state": "A task decomposition sequences approvals before the technical check; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, during the quarterly review. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9488f56f4cbc6921_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.08, 0.4, 0.51, 0.01], "state": "An agent plan skips the sandbox validation step; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_797ad98a05e1e0c9_n", "kind": "noul", "options": ["false", "true"], "target": [0.7, 0.3], "state": "An agent plan parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e0792432ad7f0fd8_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.52, 0.33, 0.04, 0.11], "state": "An automation plan sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: for a compliance-critical workflow, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7b03835b56b94d32_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A proposed workflow parallelizes steps that share a lock; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, shortly after a software update. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_298eb9f451829a86_n", "kind": "noul", "options": ["false", "true"], "target": [0.45, 0.55], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: with a hard spend cap, with a new vendor onboarded. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_38862ae2ba1f7e57_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition assumes an API quota that expired last month; additionally sequences approvals before the technical check. Context: with a hard spend cap, during an external partner integration. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_aba648158aa1a6a9_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.13, 0.84, 0.03], "state": "A proposed workflow skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e6e2936559b2bcd5_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "An automation plan handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the review queue.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b979252c26119420_n", "kind": "noul", "options": ["false", "true"], "target": [0.73, 0.27], "state": "A task decomposition parallelizes steps that share a lock; additionally skips the sandbox validation step. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_2edbfa54a492ca18_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during the quarterly review. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0a7cde69c202d5fd_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.01, 0.02, 0.15, 0.68, 0.09], "state": "An agent plan ends with a human confirmation gate; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_01cd0da562c8c559_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.08, 0.0, 0.05, 0.87], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during the morning shift. Reported by the monitoring system.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_88f9c0b0eb9b5f0d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.34, 0.58, 0.02], "state": "A task decomposition calls the refund API before checking dispute status; additionally skips the sandbox validation step. Context: with a 40-minute deadline, under reduced staffing. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9a7d9cd7eb64c642_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.03, 0.3, 0.6, 0.03], "state": "An automation plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a hard spend cap, during the holiday freeze. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_747e047521520a7d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.05, 0.2, 0.62, 0.1], "state": "An automation plan includes a rollback step for each mutation; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d437029cd105254f_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.13, 0.19, 0.43, 0.16, 0.02], "state": "A task decomposition parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b767c931f026433e_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5b637ac7d0986ce6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.08, 0.49, 0.36, 0.01], "state": "An agent plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0231052b5cedaf4a_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.26, 0.71, 0.02], "state": "An automation plan requests write access it never uses; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_20a3331b7a65035c_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.5, 0.45, 0.04], "state": "A proposed workflow parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, during peak load. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_654a2f51e695bf13_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A task decomposition ends with a human confirmation gate; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, during the morning shift. Reported by the monitoring system.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_43305fee0c96062c_n", "kind": "noul", "options": ["false", "true"], "target": [0.9, 0.1], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_341882cab392440a_n", "kind": "noul", "options": ["false", "true"], "target": [0.5, 0.5], "state": "An agent plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_fe57706a36feb7e7_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.08, 0.02, 0.04, 0.25, 0.56, 0.05], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally requests write access it never uses. Context: with a 40-minute deadline, during normal operations. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e2ccc94a7d2d64c5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.05, 0.08, 0.04, 0.83], "state": "A task decomposition uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: under a read-only budget, with staffing at reduced levels. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_d62d1d6eb099b4b3_n", "kind": "noul", "options": ["false", "true"], "target": [0.91, 0.09], "state": "An automation plan skips the sandbox validation step; additionally parallelizes steps that share a lock. Context: with a hard spend cap, while on-call rotation changed. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1a97a1b2f37eb1b5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.01, 0.08, 0.89], "state": "A task decomposition ends with a human confirmation gate; additionally requests write access it never uses. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31ade2e6560f3af5_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.23, 0.17, 0.09, 0.51], "state": "An agent plan sequences approvals before the technical check; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during an external partner integration. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b93c73174df2fcb0_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.68, 0.3, 0.02], "state": "An automation plan calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while legacy systems are still active. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b29a490991040093_n", "kind": "noul", "options": ["false", "true"], "target": [0.61, 0.39], "state": "An automation plan ends with a human confirmation gate; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63b50a4a58f08cb5_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "A proposed workflow ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a hard spend cap, under an active deadline. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_11f1afa862ca794c_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.51, 0.07, 0.26, 0.16], "state": "An automation plan parallelizes steps that share a lock; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the system migration. Reported by the operator.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e187d6ff600f7e52_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.57, 0.29, 0.13], "state": "An automation plan parallelizes steps that share a lock; additionally estimates 2 hours for a task with a known 2-day dependency. Context: under a read-only budget, at the end of the fiscal period. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e801a98765424a7b_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An agent plan includes a rollback step for each mutation; additionally parallelizes steps that share a lock. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ab5ab0b6e54a0597_n", "kind": "noul", "options": ["false", "true"], "target": [0.87, 0.13], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally skips the sandbox validation step. Context: under a read-only budget, following a recent audit. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_69ef1e2607449942_n", "kind": "noul", "options": ["false", "true"], "target": [0.22, 0.78], "state": "A task decomposition fetches customer data before verifying the account ID exists; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_28c700f379d568ad_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "A task decomposition skips the sandbox validation step; additionally ends with a human confirmation gate. Context: with a 40-minute deadline, during peak load. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_442f16a6d9a0f15e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.15, 0.14, 0.13, 0.38, 0.17, 0.03], "state": "A task decomposition parallelizes steps that share a lock; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35dedc4e441cbf9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.14, 0.79, 0.06], "state": "An agent plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, under an active deadline. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e79e0b50df176cae_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.3, 0.44, 0.21, 0.01], "state": "A task decomposition skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: with a hard spend cap, during normal operations. Reported by the supervisor.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_60b06e1f0ebb154f_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.41, 0.33, 0.21, 0.05], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, during the holiday freeze. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_848d9d0dab8a56af_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.27, 0.55, 0.17], "state": "An agent plan skips the sandbox validation step; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the holiday freeze. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_14c387dad35e6ea5_n", "kind": "noul", "options": ["false", "true"], "target": [0.75, 0.25], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31d31d3b5532172f_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.46, 0.49, 0.04], "state": "A task decomposition uses batch endpoints where streaming is required; additionally includes a rollback step for each mutation. Context: with a hard spend cap, immediately after go-live. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_73e4b79db552ef22_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.03, 0.12, 0.47, 0.31, 0.02], "state": "A proposed workflow uses batch endpoints where streaming is required; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, during a scheduled maintenance window. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_42ffe0d486f7774b_n", "kind": "noul", "options": ["false", "true"], "target": [0.65, 0.35], "state": "An automation plan requests write access it never uses; additionally handles the failure mode explicitly. Context: under a read-only budget, as part of a compliance sweep. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_dc21f7a35e2e802b_n", "kind": "noul", "options": ["false", "true"], "target": [0.42, 0.58], "state": "A task decomposition handles the failure mode explicitly; additionally ends with a human confirmation gate. Context: under a read-only budget, during normal operations. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9d37ce6a5c5e44fb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.17, 0.71, 0.12], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_94e076f83aee26bc_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.01, 0.22, 0.76, 0.01], "state": "A task decomposition handles the failure mode explicitly; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_533d3c740b37e234_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.02, 0.15, 0.4, 0.4, 0.02], "state": "An agent plan assumes an API quota that expired last month; additionally handles the failure mode explicitly. Context: for a compliance-critical workflow, during the morning shift. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1b6bb19e6ba706f4_n", "kind": "noul", "options": ["false", "true"], "target": [0.71, 0.29], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under an active deadline. Reported by the operator.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_acb10f5da57d6139_n", "kind": "noul", "options": ["false", "true"], "target": [0.96, 0.04], "state": "An automation plan requests write access it never uses; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_786a51f0cd17b110_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.01, 0.04, 0.52, 0.41, 0.01], "state": "An automation plan requests write access it never uses; additionally assumes an API quota that expired last month. Context: with a hard spend cap, while legacy systems are still active. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_90685740014cfc0e_n", "kind": "noul", "options": ["false", "true"], "target": [0.74, 0.26], "state": "An automation plan handles the failure mode explicitly; additionally includes a rollback step for each mutation. Context: with a hard spend cap, as part of a compliance sweep. Reported by the operator.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_6eb14c7aae97de9c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.04, 0.12, 0.58, 0.2, 0.01], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_da7ec481091c60ab_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.02, 0.0, 0.13, 0.85], "state": "A task decomposition handles the failure mode explicitly; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the on-call engineer.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_652b46a3c48d6ba5_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.05, 0.16, 0.55, 0.16, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally sequences approvals before the technical check. Context: with a hard spend cap, under reduced staffing. Reported by the review queue.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bdbd27f7f472b992_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.3, 0.69, 0.01], "state": "An automation plan skips the sandbox validation step; additionally assumes an API quota that expired last month. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_154f01755949dbd1_n", "kind": "noul", "options": ["false", "true"], "target": [0.67, 0.33], "state": "A task decomposition requests write access it never uses; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_9986f2f3cfc5ca97_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.01, 0.3, 0.67, 0.02], "state": "An agent plan skips the sandbox validation step; additionally estimates 2 hours for a task with a known 2-day dependency. Context: with a hard spend cap, while legacy systems are still active. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5fb5ac91f58dcfd8_n", "kind": "noul", "options": ["false", "true"], "target": [0.78, 0.22], "state": "An automation plan ends with a human confirmation gate; additionally uses batch endpoints where streaming is required. Context: with a hard spend cap, during peak load. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0139a653370db058_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.11, 0.46, 0.23, 0.2], "state": "An agent plan handles the failure mode explicitly; additionally skips the sandbox validation step. Context: under a read-only budget, under an active deadline. Reported by the on-call engineer.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5e1fe486dae0188b_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.01, 0.14, 0.51, 0.33, 0.01], "state": "An automation plan skips the sandbox validation step; additionally handles the failure mode explicitly. Context: with a hard spend cap, at the end of the fiscal period. Reported by the operator.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e5dd56afa77554dc_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.07, 0.8, 0.13], "state": "A proposed workflow skips the sandbox validation step; additionally uses batch endpoints where streaming is required. Context: for a compliance-critical workflow, mid-incident. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ac7ee7a976c48eef_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.52, 0.47, 0.01], "state": "An agent plan requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, mid-incident. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_33afe2d1a622d26d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.05, 0.29, 0.59, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7d751ac598396abb_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.52, 0.22, 0.03, 0.23], "state": "An agent plan handles the failure mode explicitly; additionally sequences approvals before the technical check. Context: for a compliance-critical workflow, while legacy systems are still active. Reported by the monitoring system.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_f84f1ad928eae97e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.04, 0.02, 0.08, 0.45, 0.39, 0.02], "state": "An agent plan uses batch endpoints where streaming is required; additionally sequences approvals before the technical check. Context: under a read-only budget, at the end of the fiscal period. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1794d9cbd14d2f7d_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.09, 0.22, 0.16, 0.41, 0.1, 0.02], "state": "An agent plan includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: under a read-only budget, under reduced staffing. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_879c368a61ddf125_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.07, 0.25, 0.37, 0.31], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally skips the sandbox validation step. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_40d9a53d498bf186_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "An agent plan fetches customer data before verifying the account ID exists; additionally requests write access it never uses. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the operator.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_3eeffff5a9955317_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.1, 0.17, 0.48, 0.16, 0.02], "state": "A proposed workflow includes a rollback step for each mutation; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during an external partner integration. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_704f2150310ff950_n", "kind": "noul", "options": ["false", "true"], "target": [0.57, 0.43], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally includes a rollback step for each mutation. Context: under a read-only budget, during a scheduled maintenance window. Reported by the monitoring system.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_31a39ba952ca88e5_c", "kind": "choice", "options": ["data_access", "external_call", "timing", "permissions"], "target": [0.0, 0.02, 0.94, 0.04], "state": "A task decomposition ends with a human confirmation gate; additionally handles the failure mode explicitly. Context: with a 40-minute deadline, at the end of the fiscal period. Reported by the supervisor.", "question": "Weakest step for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68a90a47f083c553_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.72, 0.27, 0.01], "state": "A proposed workflow fetches customer data before verifying the account ID exists; additionally sequences approvals before the technical check. Context: with a 40-minute deadline, during peak load. Reported by the supervisor.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5ff9aa6673204f93_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.11, 0.19, 0.5, 0.12, 0.01], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally ends with a human confirmation gate. Context: for a compliance-critical workflow, with staffing at reduced levels. Reported by the monitoring system.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_68f8c01fd4740486_n", "kind": "noul", "options": ["false", "true"], "target": [0.43, 0.57], "state": "A proposed workflow includes a rollback step for each mutation; additionally ends with a human confirmation gate. Context: under a read-only budget, during the holiday freeze. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_10c02a9cb53dfaaa_n", "kind": "noul", "options": ["false", "true"], "target": [0.81, 0.19], "state": "An agent plan assumes an API quota that expired last month; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, under reduced staffing. Reported by the monitoring system.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b235f6be31f14a76_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.06, 0.79, 0.15], "state": "An automation plan estimates 2 hours for a task with a known 2-day dependency; additionally uses batch endpoints where streaming is required. Context: with a 40-minute deadline, during the holiday freeze. Reported by the operator.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_800c2e193528ce37_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.0, 0.01, 0.19, 0.75, 0.04], "state": "An agent plan assumes an API quota that expired last month; additionally parallelizes steps that share a lock. Context: with a 40-minute deadline, while on-call rotation changed. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_c0f6fd65e2411611_n", "kind": "noul", "options": ["false", "true"], "target": [0.95, 0.05], "state": "A task decomposition calls the refund API before checking dispute status; additionally sequences approvals before the technical check. Context: with a hard spend cap, as part of a compliance sweep. Reported by the automated check.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_49faad74a96daf20_n", "kind": "noul", "options": ["false", "true"], "target": [0.47, 0.53], "state": "An automation plan skips the sandbox validation step; additionally sequences approvals before the technical check. Context: under a read-only budget, during normal operations. Reported by the on-call engineer.", "question": "Is this scenario one where: the plan can complete within budget?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_de0e925a5a4ea8fe_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "An automation plan requests write access it never uses; additionally ends with a human confirmation gate. Context: under a read-only budget, during an external partner integration. Reported by the supervisor.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_62409e747b7d6f5e_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.05, 0.15, 0.54, 0.23, 0.01], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: under a read-only budget, with staffing at reduced levels. Reported by the supervisor.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_e67f41eef1c700b0_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.04, 0.34, 0.57, 0.02], "state": "A proposed workflow parallelizes steps that share a lock; additionally assumes an API quota that expired last month. Context: with a 40-minute deadline, under reduced staffing. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_958bd4d1642f3f2c_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.02, 0.01, 0.01, 0.1, 0.71, 0.15], "state": "An automation plan skips the sandbox validation step; additionally calls the refund API before checking dispute status. Context: with a 40-minute deadline, during the morning shift. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_a21de5fbb244ee77_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.33, 0.49, 0.17], "state": "A task decomposition ends with a human confirmation gate; additionally skips the sandbox validation step. Context: for a compliance-critical workflow, at the end of the fiscal period. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_5391cb894b202033_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.07, 0.22, 0.24, 0.31, 0.12, 0.04], "state": "An automation plan sequences approvals before the technical check; additionally parallelizes steps that share a lock. Context: for a compliance-critical workflow, as part of a compliance sweep. Reported by the on-call engineer.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bad182e4638262c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.04, 0.16, 0.55, 0.2, 0.02], "state": "A proposed workflow sequences approvals before the technical check; additionally includes a rollback step for each mutation. Context: with a 40-minute deadline, with staffing at reduced levels. Reported by the review queue.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_ea2d938aeec9c4b1_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.01, 0.02, 0.15, 0.73, 0.06], "state": "A task decomposition parallelizes steps that share a lock; additionally calls the refund API before checking dispute status. Context: for a compliance-critical workflow, while on-call rotation changed. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_0c7ec41836541945_n", "kind": "noul", "options": ["false", "true"], "target": [0.79, 0.21], "state": "An automation plan handles the failure mode explicitly; additionally fetches customer data before verifying the account ID exists. Context: with a hard spend cap, during a scheduled maintenance window. Reported by the on-call engineer.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_87e70297987fc237_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.05, 0.14, 0.22, 0.44, 0.13, 0.02], "state": "An automation plan includes a rollback step for each mutation; additionally fetches customer data before verifying the account ID exists. Context: under a read-only budget, shortly after a software update. Reported by the automated check.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_b1d1706f3b1b73c6_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.01, 0.03, 0.14, 0.46, 0.33, 0.03], "state": "A proposed workflow handles the failure mode explicitly; additionally requests write access it never uses. Context: with a hard spend cap, immediately after go-live. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_35eb8a38e01d39fa_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.01, 0.83, 0.14, 0.02], "state": "A task decomposition includes a rollback step for each mutation; additionally requests write access it never uses. Context: with a 40-minute deadline, during the quarterly review. Reported by the automated check.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_7ba37c74bed37a0d_n", "kind": "noul", "options": ["false", "true"], "target": [0.72, 0.28], "state": "A task decomposition sequences approvals before the technical check; additionally assumes an API quota that expired last month. Context: for a compliance-critical workflow, during an external partner integration. Reported by the review queue.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_8d1b72dbe90212ef_c", "kind": "choice", "options": ["approve_plan", "revise", "reject", "defer_to_human"], "target": [0.0, 0.48, 0.27, 0.25], "state": "A proposed workflow estimates 2 hours for a task with a known 2-day dependency; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, under reduced staffing. Reported by the review queue.", "question": "Verdict for this scenario.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_1893c5ae8a303aa8_n", "kind": "noul", "options": ["false", "true"], "target": [0.05, 0.95], "state": "A proposed workflow requests write access it never uses; additionally includes a rollback step for each mutation. Context: under a read-only budget, during the system migration. Reported by the automated check.", "question": "Is this scenario one where: a step violates least privilege?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_bd5087b9da3ae256_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.03, 0.03, 0.1, 0.32, 0.48, 0.04], "state": "A task decomposition calls the refund API before checking dispute status; additionally includes a rollback step for each mutation. Context: with a hard spend cap, following a recent audit. Reported by the on-call engineer.", "question": "Rate completion likelihood for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_85327ca3b3928bfa_n", "kind": "noul", "options": ["false", "true"], "target": [0.8, 0.2], "state": "An automation plan uses batch endpoints where streaming is required; additionally requests write access it never uses. Context: for a compliance-critical workflow, during the system migration. Reported by the review queue.", "question": "Is this scenario one where: the ordering is correct?", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"} {"id": "v3_63d2c885cbee05bf_s", "kind": "score", "options": ["0", "1", "2", "3", "4", "5"], "target": [0.0, 0.0, 0.04, 0.32, 0.62, 0.02], "state": "An automation plan uses batch endpoints where streaming is required; additionally fetches customer data before verifying the account ID exists. Context: for a compliance-critical workflow, with a new vendor onboarded. Reported by the automated check.", "question": "Rate plan quality for this scenario on a 0-5 scale.", "domain": "plan_quality", "family": "agent", "source": "yuri_v3"}