Hey Product Hunt 👋 I built grimoire because AI assistants have a consistency problem. They know best practices — they just don't apply them reliably across sessions. There's no enforcement layer. grimoire treats practices as versioned packages you install into your AI agents. Declare what you want in `grimoire.toml`, run `grimoire install`, and your AI agents now have a shared, versioned practice library. Then run `grimoire check` to audit your project against those practices — structured JSON output, HTML dashboard, GitHub Actions integration. **What makes it different:** - **Skills as packages** — fork any registry, publish your team's internal practices as a private git repo, pin versions - **Compliance engine** — continuously audits projects, not just individual files; threshold-based CI integration - **MCP server** — AI agents can self-manage their own skill library via tools - **LSP server** — real-time compliance diagnostics pushed to your editor gutter on every save - **Cross-agent** — Claude, Codex, Gemini, Antigravity, OpenCode, OpenClaw One-line install. Works on macOS, Linux, and Windows. → https://grimoire.jeffreytse.net Happy to answer questions about the architecture — especially the compliance engine and how skill authoring standards work.

Grimoire

A package manager for AI agent best practices

A package manager for AI agent best practices