Upload graph.jsonl with huggingface_hub

graph.jsonl

@@ -26,6 +26,7 @@
 {"@id":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0018-record-downstream-applications-as-first-class-context.md","@type":["ADR","TechArticle"],"name":"ADR-0018: Record Downstream Applications as First-Class Context","description":"The start-of-program repository did not know its own downstream: Contemplative Agent's description was past-direction only (upstream substrate), and four DOI-registered downstream repositories were unmentioned — surfaced as observed pain when the position paper could not cite AKC as its own primary source. Records the corrective: downstream applications are first-class context. Contemplative Agent's role is two-way (substrate for ADR-0002–0005 and operational re-implementation running the six-phase cycle over its own episode logs, every promotion human-gated, demonstration ongoing); crystallized research lines are recorded with relationship facts and DOIs only, importing no downstream content (mechanism-only rule preserved). Itself the record of a Maintain-phase execution on the repository's own self-description."}
 {"@id":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0019-cycle-structure-is-provisional.md","@type":["ADR","TechArticle"],"name":"ADR-0019: The Cycle's Structure Is Provisional — Skills, Bindings, and Phases Held Lightly","description":"AKC holds its own structure provisionally: no layer of the cycle's self-description — skills, the phase-to-skill binding, or the six-phase set — is a fixed essence. Operationalizes the Emptiness axiom (contemplative-axioms, Laukkonen et al. 2025) for AKC's own frame, correcting a reification error in which the 'bijective' binding was read as an invariant that forbade a second Curate skill. Three concrete moves: (1) the binding is reworded across graph.jsonld, the README phase table, and llms.txt from 'bound bijectively to X' to 'currently scaffolded by X (a mutable snapshot)'; (2) Curate is layered — its structural / code layer is skill-health (a dangling-reference, missing-artifact scan; an ADR-0008 guard, deterministic), its semantic / judgment layer is skill-stocktake — enumerate-then-decide, skill-health published as a standalone repository and referenced like the other cycle skills, its SkillOps four-dimension rubric staying content-side; (3) the skill set both grows (skill-health joins) and dissolves (skill-comply's compliance-checking is a candidate for substrate absorption by the host harness, recorded experimental n=1, not retired). The six-phase count is de-reified but not removed — provisional is not arbitrary; the structures remain the load-bearing current articulation, held lightly. The ADR is itself provisional. Extends ADR-0009, applies ADR-0008, bounded by ADR-0011, continues ADR-0018, grounded in contemplative-axioms (Emptiness).","groundedIn":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0009-akc-is-a-cycle-not-a-harness.md"}
 {"@id":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0020-readme-minimal-floor.md","@type":["ADR","TechArticle"],"name":"ADR-0020: Minimal-floor README with Single-location Theme Presentation","description":"Amends ADR-0012's README-side prescription after its fix overshot: front-loading the three core themes produced a double presentation (a four-paragraph 'What is AKC?' plus a three-subsection 'Why AKC') and a skill list spread across three tables, growing the README to 302 lines / 8 tables — observed as operator pain on 2026-07-07. The corrective: README.md moves to a minimal floor — the three themes get exactly one full presentation (in 'Why AKC', fixed order: cognitive resource → intent alignment → the cycle changes the human too), the skill list collapses to the single phase→skill table, and design principles, references, the harness-engineering comparison, and adoption levels become pointers into the existing machine-readable floor (llms-full.txt, individual ADRs) rather than second copies. ADR-0012's front-load principle (theme phrases in the first thirty lines, themes before mechanism) and its llms.txt / llms-full.txt commitments remain in force unchanged; llms-full.txt gains explicit responsibility as the canonical information floor. Second correction in the same front-door alignment family: ADR-0012 corrected invisibility, ADR-0020 corrects the overshoot into repetition. Same spirit as ADR-0019 applied to the document layer — the README's shape is a provisional structure, revised on observed pain and recorded rather than silently patched.","groundedIn":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0012-front-load-three-core-themes.md"}
+{"@id":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0021-replace-growth-tagline.md","@type":["ADR","TechArticle"],"name":"ADR-0021: Replace the Growth Tagline","description":"Replaces the tagline 'A knowledge cycle for AI agents — one that grows with the people who shape it' (introduced by ADR-0009, preservation reaffirmed by ADR-0010 and ADR-0012) with 'A knowledge cycle for AI agents — agent behavior compounds, human judgment sharpens.' The earlier phrasing collided with the tagline of a widely adopted agent-centric autonomous self-improvement product ('The agent that grows with you', February 2026), which predated it by two months — a semantically inverting collision: that product's 'grows' is ungated agent-side self-improvement, precisely the framework class AKC positions against, so the surface similarity read AKC as derivative and obscured its human-gated, bidirectional core. The new parallel two-clause form abandons the shared relative-clause-with-'grows with' structure, encodes bidirectionality in the syntax itself (one clause per end of the loop), and carries theme 1 (human judgment) and theme 3 (bidirectional loop) simultaneously; 'judgment sharpens' was already canonical AKC vocabulary (coin-sparingly / anchor-densely). Supersedes only the tagline-preservation clauses of ADR-0010 and ADR-0012; the cycle-not-harness repositioning (ADR-0009), the three-theme front-loading discipline, and the ADR-0012/0020 Measure checklists stand unchanged.","groundedIn":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0009-akc-is-a-cycle-not-a-harness.md"}
 {"@id":"https://shimo4228.github.io/shimo4228/vocab#concept/six-phase-loop","@type":["Concept","DefinedTerm"],"name":"six-phase loop","description":"Bidirectional growth loop running Research, Extract, Curate, Promote, Measure, Maintain. Each phase is currently scaffolded by composable skills — a mutable snapshot, not a fixed bijection (ADR-0019); the cycle stays stable even as individual skills are added, layered, or dissolved.","groundedIn":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0009-akc-is-a-cycle-not-a-harness.md"}
 {"@id":"https://shimo4228.github.io/shimo4228/vocab#concept/three-layer-structure","@type":["Concept","DefinedTerm"],"name":"three-layer structure","description":"AKC's stacked architecture: principles (ADRs), patterns (design-pattern skills), implementation (composable skills). Separating layers decouples rate of change.","groundedIn":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0003-three-layer-distillation.md"}
 {"@id":"https://shimo4228.github.io/shimo4228/vocab#concept/scaffold-dissolution","@type":["Concept","DefinedTerm"],"name":"scaffold dissolution","description":"Property that explicit AKC skills can be dropped once the cycle has been internalized. The skills are scaffolding, not the goal. Dissolution is the intended end state, not a fallback.","groundedIn":"https://github.com/shimo4228/agent-knowledge-cycle/blob/main/docs/adr/0009-akc-is-a-cycle-not-a-harness.md"}