Agent rules reviewer

    1

    Structural and content review for agent instruction files (CLAUDE.md, AGENTS.md, .mdc)

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Agent rules reviewer

    Example session with this skill installed

    Review my .mdc rule for my typescript-react project. I'm worried it's too long and might not be activating correctly on my .tsx files.

    • Read your context and instructions
    • Compiled the agent rules reviewer

    Illustrative audit output for the example configuration and project details below, not a recorded customer session.

    I have completed the structural and content audit of your .mdc rule. I found a mismatch in your glob patterns and a rough size estimate well above my reference sample.

    Agent Rule Audit Report

    CategoryFindingSeverity
    Structureglobs: ["src/*.ts"] does not match your .tsx files.Structural Problem
    ActivationalwaysApply: false with globs allows file-based auto-attachment. A description is not required for that mode.Informational
    Size~3,200 estimated tokens (roughly 1.6% of a 200k window).Review Item
    ContentNo Agent Governance section (no rules on branching/secrets).Content Gap
    WordingUses "Please consider using..." for core linting rules.Wording Hedge

    Size & Cost Estimate

    • Estimated tokens: 3,200 (rough chars/4 estimate from ~12,800 characters).

    Reference point: Well above the median of 348 estimated tokens in a November 2026 sample of 12 root CLAUDE.md files from major repositories. That sample is small and not an industry benchmark.

    • Session cost: At an illustrative $15 per 1M input tokens, roughly $0.048 per new session start.

    Content Pillar Check

    1. Orientation: Present. Describes the React/Vite layout.
    2. Operations: Present. Includes npm run dev and vitest.
    3. Governance: MISSING. No instructions on PR standards or API key safety.
    4. Pointers: Present. Links to internal UI kit documentation.

    Next steps

    1. Update globs to ["src/**/*.{ts,tsx}"] so the rule activates on your React components.
    2. Add a description only if relevance-based selection is also intended; it is not needed for matching-glob activation.
    3. Rewrite hedged rules in imperative language (for example, "Use functional components" instead of "Consider using...").
    estimated tokens in a November 2026 sample of 12 root CLAUDE.md

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Version notice: Agensi lists corrected v1.2 as the latest approved package. Earlier v1.0 guidance incorrectly said a missing description makes a Cursor rule manual-only. Matching globs can auto-attach without a description. The delivered package contents have not been independently re-read through an agent connection; apply this correction if you encounter the older wording.

    Instruction files like CLAUDE.md or Cursor .mdc rules can grow bloated, fail to activate, or contain conflicting directives. This skill reviews agent instruction files for structural problems, activation-mode mistakes, size cost, and content gaps, and separates real problems from review prompts.

    What it does

    • Structural validation flags .mdc frontmatter issues, missing descriptions when relevance-based selection is intended, and glob patterns worth checking against your repository structure.
    • Activation auditing applies Cursor's documented rule semantics: alwaysApply: true means description and globs are ignored for activation; with alwaysApply: false, matching globs auto-attach a rule without needing a description; a description alone supports relevance-based selection; without either globs or a description, the rule is manual-only.
    • Size estimate gives a rough token count (chars/4, labeled as an estimate) and frames it as share of the context window loaded every session.
    • Content gap analysis maps the file against four jobs instruction files typically serve: orientation, operational discipline, agent governance, and pointers to deeper docs.
    • Wording review flags hedged or soft wording in rules meant to be binding, as prompts for your review, not as verdicts about agent behavior.

    How it works

    1. File identification treats a file as a Cursor rule only when its frontmatter contains .mdc-specific keys (description, globs, alwaysApply), which avoids false failures on plain markdown files that open with YAML-style metadata.
    2. Structural check reviews activation metadata and globs against the repo tree when the tree is available.
    3. Size framing reports the estimate next to a measured reference point: in a November 2026 sample of 12 root CLAUDE.md files from major repositories, the median was 348 estimated tokens and three files exceeded 2,000. This is a small convenience sample, not an industry benchmark.
    4. Grouped findings separate structural problems (things that affect whether a rule activates at all) from things to review (judgment calls).

    Works with

    CLAUDE.md, AGENTS.md, and Cursor .mdc files.

    Use cases

    • Check why a Cursor rule is not auto-attaching to your files.
    • Review a long CLAUDE.md before committing it.
    • Sanity-check a new instruction file for structure and coverage.

    Known limitations

    Token counts are rough estimates (chars/4); tokenizers vary. Glob validation requires access to the repository file tree. Wording flags are review prompts, not verdicts.

    Token estimates are approximations based on character count. Glob pattern validation requires the agent to have permission to list or read the repository file structure.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed today

    • Free with an Agensi account

    Permissions

    Read Files
    Listedtoday

    What's inside

    Frequently Asked Questions

    This skill audits CLAUDE.md, AGENTS.md, and Cursor .mdc files for structural problems (frontmatter and glob issues that can stop a rule from activating), rough size cost, and content gaps. It flags findings for your review; you decide what to change.

    It works with plain markdown too, but it treats a file as a Cursor rule only when the frontmatter contains .mdc-specific keys (description, globs, alwaysApply), so ordinary markdown files do not produce false failures. CLAUDE.md and AGENTS.md files get structure, size, and content review.

    No complex setup is required; simply provide the content of your instruction files and your repository's file structure to receive a comprehensive audit report.

    Token counts are rough estimates (chars/4) and labeled as such; tokenizers vary. Size is framed against a November 2026 sample of 12 root CLAUDE.md files from major repositories, a small convenience sample, not an industry benchmark. Session cost figures are illustrative.

    The skill maps your file against four jobs instruction files typically serve: orientation, operational discipline, agent governance, and pointers to deeper docs. Gaps are reported as review prompts, not verdicts about agent behavior.

    New in Code Quality & Review