⚡ Instant Tools

Prompt & Rules Linter

Score system prompts, AGENTS.md, CLAUDE.md, and Cursor rules for clarity, safety, and contradictions. Free, private, browser-local.

Privacy: Your prompt never leaves this browser. Heuristic lint only — no model API call.

How to Use

  1. 1Paste your system prompt or agent rules file (AGENTS.md, CLAUDE.md, .mdc).
  2. 2Skim the grade and dimension bars — fix Consistency and Safety first.
  3. 3Replace vague phrases with concrete constraints and an output format.
  4. 4Re-lint, then download the markdown report for your PR checklist.

What gets flagged

Vague filler

“Be helpful / as needed / best practices”

Swap for measurable rules and forbidden actions.

Contradictions

“Always be thorough” + “Always be concise”

Pick one default and state exceptions explicitly.

Tips

  • •Put hard safety and “do not” rules near the top — models overweight early tokens.
  • •One rule per bullet beats paragraphs of overlapping advice.
  • •For Cursor .mdc files, set description + globs or alwaysApply.
  • •Pair with the Cursor Rules Generator when you need a clean starter file.

Frequently Asked Questions

What does the Prompt & Rules Linter check?

It scores clarity, structure, safety boundaries, contradictions, and actionability (output format, examples, tool policy). It works on system prompts, AGENTS.md, CLAUDE.md, Cursor .mdc rules, and .cursorrules.

Is this an LLM judging my prompt?

No. Checks are deterministic heuristics in your browser—fast, private, and repeatable. Use it as a pre-flight checklist before refining with a model.

Does my prompt leave the browser?

No. Lint runs 100% client-side. Instant Tools never uploads your prompt or rules file.

How is this different from the Cursor Rules Generator?

The generator creates starter rules from stack presets. This linter audits what you already wrote (or generated) for vagueness, conflicts, and missing safety.

Will a high score guarantee good agent behavior?

No. A strong score means fewer obvious failure modes. Real quality still needs evals on your tasks and models.

Is it free?

Yes. No signup, no limits, and the markdown report has no watermark.

Related Tools

Cursor Rules Generator

Generate .mdc, AGENTS.md, and CLAUDE.md

MCP Tool Checker

Score agent tool schemas for readiness

JSON to Schema

Infer Zod schemas for structured outputs

LLM Cost Calculator

Estimate GPT · Claude · Gemini API spend

OpenAPI to Tool Schemas

Turn APIs into function-calling schemas

Prompt & Rules Linter

Score system prompts and agent rules for clarity