Skip to content

Keep your skills sharp.

skillsaw lints the files that steer your AI coding agents: skills, plugins, instructions, and tool configs across Claude Code, Codex, Copilot, Agent Skills, Google Antigravity, OpenCode, and more. It catches security risks, structural flaws, and unclear instructions with 110 rules, then applies deterministic autofixes.

PyPI version Tests License skillsaw grade

Get Started View Rules

  • 🛡 Security & Supply Chain

    Supply chain protection blocks dangerous lifecycle hooks (curl | sh, eval, arbitrary execution), prohibited MCP servers, invisible Unicode (ASCII smuggling), high-entropy encoded payloads, and unallowlisted dynamic context before they run.

  • 📦 Multi-Ecosystem Structure

    Schema and syntax validation across supported ecosystems: Agent Skills, Claude Code, OpenAI Codex, Grok Build, Google Antigravity, native OpenClaw plugins, Agent Plugins v1, Copilot custom agents, OpenCode, Muse Code, Pi, APM, and MCP Registry.

  • 🔧 Deterministic Autofixes

    Safe, instant autofixes via skillsaw fix; bundled skills guide coding agents through remaining manual resolutions — see demo below.

  • 🧠 Context & Token Economy

    Research-backed content checks flag instruction drift, excessive instruction budgets, long sections, contradictions, and redundant tooling.

  • 🤖 CI-Ready & Baselines

    GitHub and GitLab integration with PR comments, SARIF reporting, and baselines for gradual, zero-friction adoption.

  • 🔌 Extensible

    Custom rules in Python, pip-installable rule plugins, and typed lint tree traversal tailor skillsaw to your repository.

Reading this with an AI agent? Fetch llms.txt for an index of the docs, or llms-full.txt for everything in one plain-markdown file.