What problem does it solve? Repeatedly correcting an AI coding agent wastes effort because the corrections never become permanent rules. This Skill turns mined, judge-accepted correction candidates into typed, tested Python hooks and opens one pull request per candidate with verbatim evidence. ## Core Features & Use Cases - Candidate verification: Enumerates threshold-eligible candidates via the capt-hook review CLI and re-verifies every quoted correction verbatim against session transcripts before acting. - Routing and overlap screening: Screens each create candidate against active repo-local and pack hooks, deciding between a new hook, an EXTEND edit, a FIX amendment, or a duplicate skip, and routes the PR to the correct repository. - Drafting and PR creation: Delegates hook drafting to the authoring-hooks skill, verifies with capt-hook test in a worktree or clone, and opens exactly one PR per candidate with an evidence-backed body. - Use Case: After a session where a user repeatedly corrected force-push behavior, the detached reviewer spawns this Skill, which drafts a block_command hook, proves it with inline tests, and opens a PR citing the verbatim corrections. ## Quick Start Invoke /captain-hook:scanning-sessions --transcript with the path to the ended session's transcript inside the watched repository.
