Agent Lint is an open-source linter for AI coding-agent configuration. It validates Claude Code, Cursor, Codex, AGENTS.md, agent skills, plugins, and MCP configuration with ~300 rules. Built in Rust, it supports configurable severity and suppression, targeted checks, GitHub Actions integration, and prebuilt binaries for Linux and macOS.
Howdy folks 👋
I built Agent Lint because coding-agent configuration is starting to feel like a real software surface of its own.
Claude Code, Cursor, Codex, MCP, AGENTS.md, skills, plugins, and project instructions are getting increasingly powerful, but they’re also becoming easier to misconfigure in subtle ways. I wanted the equivalent of a compiler/linter sanity check for this new layer of the stack.
Agent Lint now has roughly 300 rules and is fully open source.
A fun detail: the project itself was built 100% with Claude, including architecture, implementation, tests, refactoring, documentation, and much of the rule design. So Claude helped build a tool that now lints Claude configuration. 🙂
I’d especially love feedback on:
• configuration patterns I’m missing
• rules that are too strict or too lenient
• false positives
• additional Claude Code, Cursor, Codex, MCP, or AGENTS.md checks that would be useful
GitHub issues and contributions are very welcome. Thanks for taking a look!