What problem does it solve? When work must pause or move to a new AI-agent session, context is lost and the next session may fabricate missing state, overwrite unrelated notes, or treat stale authorization as current. This Skill produces a self-contained, evidence-grounded handoff so a blank-slate session can resume safely. ## Core Features & Use Cases - Evidence classification: Separates observed, reported, inferred, unknown, and conflicting claims, and tracks confirmed versus pending or superseded decisions. - Orthogonal state reporting: Assigns exactly one handoff state (Ready to resume, Needs confirmation, Blocked) and one persistence state (Written, Draft only, Not written). - Safe persistence rules: Writes only to an explicitly authorized destination, refuses to overwrite newer or different-task handoffs, strips secrets and private data, and never renews prior high-risk authorization. - Use Case: You are pausing a parser refactor mid-task. The Skill records the confirmed retry decision, the unverified test report, the uncommitted diff, and the next safe action with entry and stop conditions as a draft, since no storage destination was authorized. ## Quick Start Ask the agent to record a session handoff for the current task so a later session can resume it safely.
