claudelint
A Node CLI that lints Claude Code project configuration and ships as a plugin, so agents can validate their own setup.
A claudelint run against a deliberately broken project, surfacing every category's problems at once. Plain text by default; JSON or SARIF for CI pipelines and agent loops.
I started working on claudelint shortly after Anthropic launched Skills in Claude Code. Without a validation step in the loop, Claude wrote Skills that didn't trigger when they should have, or ran and produced broken output. These types of failure didn't surfaces as errors. Claude Code's /doctor checks runtime health, but nothing was checking the configuration tree itself. I wanted Expo Doctor for Claude Code projects.
claudelint validates the entire configuration tree, including how files cross-reference each other. Output is plain text by default, with --format json and --format sarif for CI and agent loops.
claudelint also installs as a Claude Code plugin, so validation runs inside the agent's loop. Claude can validate a project before applying changes, then iterate against the output. The docs site is written so an agent can set the tool up cold, on a fresh project, without a human in the loop. When Claude makes the same mistake twice, the fix isn't a longer CLAUDE.md, it's a custom rule that works exactly like you'd expect it to.