rxa3c/chat2skill
Extracting and iterating skills from daily conversations with AI
Learns from your own conversations: after a session it looks for corrections, preferences, constraints and project facts, distils them into atomised SKILL.md files and a project memory store, and injects the relevant ones into later sessions. Candidates that a judge rejects are fed back into generation up to two times, and a maintenance pass merges near-duplicates and archives weak skills so the prompt does not grow forever.
Install
⚠️ Upstream documents a local-checkout install — the command it gives is a path on your own machine, so there is nothing pasteable here. See the README for its exact steps.
Compatibility
Domain-general but honest about its target: it works best with Claude Code, Codex and Cursor, and for DSH it runs through the lifecycle hooks — agent/pre-step for retrieval and agent/turn-stopping for the shared guard and learning. Its hooks run Python, so the machine running the agent must have Python available, and the plugin cannot initialise local storage or run retrieval without it. Memory and skills are local: a SQLite database for project memory plus SKILL.md files, with a synthesised PROJECT_SKILL.md for human review. The LLM path uses proposer, generator and judge stages, and when no LLM is available the loop still runs on keyword detection and template generation. Extraction triggers only when a correction or constraint signal appears, not on every session, and long sessions are trimmed to the latest analysis window.
Details
- Repo: rxa3c/chat2skill
- Category: Agent Capabilities
- Stars: 2
- Version: not published to npm; installs from a checkout
- Last push: 2026-08-27
- First seen: 2026-06-10
Alternatives
MJorgin/dsh-media-skills · xiaoxiaosrm/dsh-mattpocock-skills · paean-ai/8x-skills