AI rules sanity check
Paste a Cursor .mdc rule, a CLAUDE.md, or an AGENTS.md below. This runs structural checks that catch the most common reasons an assistant silently ignores a rule file. It is a sanity check, not a full linter - and it runs entirely in your browser.
Runs entirely in this tab. Close it and your text is gone.
What this checks (and what it can't)
The frontmatter and glob checks are structural - verifiable from the file's text against Cursor's documented rule types (Always Apply, Apply Intelligently, Apply Manually, and glob auto-attach). The size, contradiction, and hedged-wording checks are heuristics: they surface lines for your review and can be wrong. In our own 12-repo survey, both contradiction flags were false positives - treat heuristic warnings as review prompts, not verdicts. Whether your rules are good instructions is a judgment this page doesn't make. Claude Code loads CLAUDE.md at session start; AGENTS.md is the cross-tool convention adopted by several agents. Token estimates use the rough 4-characters-per-token average and can be off in either direction.