Living Memory then Proactive Execution: prepare, refresh, return
Living Memory helps an agent carry a Session forward, but Proactive Execution still needs a clear boundary before it prepares the next step. These fictional cases make three outcomes explicit: prepare_for_review when current, scoped information supports a draft; refresh_context when a source, date, scope, or decision is stale or incomplete; and return_to_person when new access, material judgment, or an external commitment belongs with a person. The dataset is a small synthetic teaching reference, not a benchmark, product telemetry, or authorization to act. I work on Klik (pre-launch), an app-first proactive-AI direction built around deliberate Sessions, persistent Context, and permission-aware follow-through. https://pre.hiklik.ai/?utm_source=huggingface&utm_medium=dataset_discussion&utm_campaign=kickstarter_prelaunch&utm_content=proactive_execution_context_review_cases_bridge_20260902