What problem does it solve? After a slice of an implementation plan is coded, it is unclear whether the work is truly done, done with follow-ups, or still blocked. This Skill makes that closure decision explicit, distinguishes slice completion from full task completion, and persists the verdict as reviewable notes in slice files and TASK_STATE.md so progress survives across sessions. ## Core Features & Use Cases - Closure verdict classification: Classifies the current slice as ready to close, ready to close with follow-ups, or not ready to close, requiring recorded command output as evidence for each exit criterion. - Auditable state persistence: Writes a canonical five-section update to TASK_STATE.md with tamper-evident transaction headers and a JSONL verification log entry per section. - No-op and checkpoint handling: Returns a no-op when closure would not materially change memory, and records resumable checkpoints when a single pending verdict (such as a human feel-verdict) blocks closure. - Use Case: After implementing a HIGH-complexity slice whose exit criteria need manual verification, run this Skill to validate the evidence, close the slice, and route to the next step such as sync-task-state or implement-slice-complement. ## Quick Start Ask the AI to run slice-closure on the active task to decide whether the just-implemented slice can be closed and to update TASK_STATE.md with the verdict.
