Поиск решений и паттернов по прошлым сессиям

productivity

recall-reasoning is a Claude Code skill that searches previous work sessions to surface relevant decisions, successful approaches, and documented failures — so past learnings directly inform current work. It queries two sources: the Artifact Index (handoffs, plans, and ledgers with post-mortem sections) via `artifact_query.py` supporting `--outcome SUCCEEDED/FAILED` filters, and Reasoning Files stored as `reasoning.md` inside `.git/claude/commits/*/`, which capture build attempts, test failures, and commit context. Results are tagged ✓/✗/? so patterns are immediately actionable. Not user-invocable directly, this skill is triggered automatically when starting work similar to past sessions, investigating why a decision was made, or debugging a previously encountered issue.