skill-sedimentlisted

> - **Author**: Evan Song <Songhonglei@users.noreply.github.com> > - **Repository**: https://github.com/Songhonglei/build-better-skills > - **License**: MIT — see [LICENSE](./LICENSE) > - **Version**: 1.0.2 # skill-sediment One-shot installer for the **OpenClaw skill-sediment plugin extension** — auto-distill successful conversation workflows into `SKILL.md` files, then promote them to `skills/`. ## Core Facts - **What is being ported**: an OpenClaw plugin extension (not an MCP, not a standalone service) that listens to conversations and writes successful workflows as SKILL.md. - The plugin is ≈340KB of pure TypeScript source (9 files), depending only on OpenClaw core's built-in `plugin-sdk` + `typebox`. - **No embedding model required** — sediment uses LLM semantic judgement, not vector retrieval. - The plugin source is **already embedded in `assets/plugin-source/`** (flat directory of 9 TypeScript files + `package.json` + `openclaw.plugin.json`) — offline install works out of the box. A legacy `.tar.gz` fallback path is also supported for backward compatibility. ## Quick Install ```bash cd <this skill directory> python3 scripts/manage.py install # Then follow the prompt to manually restart: openclaw gateway restart # Verify after restart: python3 scripts/manage.py doctor ``` In most cases `plugins.allow` is empty, so the plugin loads automatically from the convention directory — **no config change required**, just restart. `--valid-agent-id` is only needed