{"skill":{"slug":"taku-reflect","displayName":"Taku Reflect","summary":"User-invoked reflection. Three modes: Learn (script-backed recording, searching, pruning, exporting, and optional bootstrap for user-approved patterns, pitfa...","tags":{"latest":"1.0.0"},"stats":{"comments":0,"downloads":77,"installsAllTime":0,"installsCurrent":0,"stars":0,"versions":1},"createdAt":1778495106348,"updatedAt":1778495809038},"latestVersion":{"version":"1.0.0","createdAt":1778495106348,"changelog":"taku-reflect 1.0.0\n`/taku-reflect` is manual by design. Only user-approved patterns, pitfalls, preferences, and discoveries get preserved. .taku/learnings/*.jsonl is canonical, managed by the bundled reflect scripts/learnings.py.\n\n- Initial release introducing three reflection modes: Learn (capture, search, prune, export user-approved learnings), Retro (evidence-based weekly retrospectives via git/metrics analysis), and Write Skill (codify recurring learnings into reusable skills).\n- Learn mode uses script-backed operations, storing structured learnings in project-scoped JSONL files and supporting optional project bootstrap with protocol suggestions for AGENTS.md/CLAUDE.md.\n- Manual invocation only: learnings are never recorded, updated, or pruned without explicit user action.\n- Supports targeted reflection workflows triggered by specific phrases or natural reflection moments after work completion.\n- All outputs organized with clear sections for new learnings, relevant past entries, promotion suggestions, and bootstrap status.","license":"MIT-0"},"metadata":null,"owner":{"handle":"kkenny0","userId":"s17e5sm9mfd2xfq0bhb9w9bd9n840vsx","displayName":"KennyWu","image":"https://avatars.githubusercontent.com/u/18048025?v=4"},"moderation":null}