Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Agent config auditor

    1

    Audit Claude Code and Cursor configurations for shallow skills, missing hooks, and ambiguous MCP tool descriptions.

    Free

    Security scanned
    Agent config auditor

    Agent config auditor

    Example session with this skill installed

    Run a config audit on the current project to check for shallow skills and missing safety hooks.

    • Read your context and instructions
    • Compiled the agent config auditor

    I have audited the configuration directories for this project. The scan identified 4 issues across your skill definitions and tool settings.

    Agent Config Audit Report

    SeverityFile PathIssue TypeDescription
    High.claude/settings.jsonMissing HookNo PreToolUse hook defined. Destructive tools will execute without confirmation.
    Mediumskills/deploy-bot/SKILL.mdShallow SkillFile contains instructions but lacks a scripts/ folder or tool references.
    Medium.cursor/mcp.jsonAmbiguous MCPTool database-query has a vague description. Increase specificity to avoid hallucinations.
    Low.cursor/mcp.jsonMissing EnvServer stripe-api has no environment variables declared in config.

    Next steps

    1. Add a PreToolUse hook to .claude/settings.json to guard rm and git push operations.
    2. Convert deploy-bot from a text-only skill into an executable workflow by adding a handler script.
    3. Update the description for the database-query tool in your MCP config to include specific schema constraints.

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

    About this skill

    Configuring AI agents often results in shallow skills that act as static text rather than executable workflows. This skill audits your local agent configuration to identify missing logic, unsafe tool hooks, and ambiguous server definitions.

    What it does

    • Identifies shallow skills by flagging SKILL.md files that lack associated scripts or tool references.
    • Audits safety hooks by checking for missing PreToolUse configurations in settings files.
    • Evaluates MCP definitions to find ambiguous tool descriptions that cause model hallucinations.
    • Validates environment variables within mcp.json to ensure servers have required credentials.
    • Pinpoints file locations by providing specific paths and line numbers for every identified issue.

    How it works

    1. Point the auditor at a project root containing .claude/ or .cursor/ directories.
    2. The script parses the markdown and JSON configuration files locally.
    3. It compares your setup against research-backed patterns for high-performance agent skills.
    4. You receive a prioritized report of configuration debt and reliability risks.

    Frameworks & tools

    This tool targets Claude Code and Cursor configurations. It is a local-first Python script with no external dependencies or network requirements.

    Why this beats prompting it yourself

    Manual reviews often miss the absence of specific hooks or the subtle ambiguity in tool descriptions that lead to runtime failures. This skill automates the detection of structural configuration smells that general-purpose prompts usually ignore.

    Use cases

    • Auditing a new agentic project for production readiness.
    • Hardening security by ensuring destructive tools have confirmation hooks.
    • Troubleshooting frequent "wrong tool" selection errors in MCP-enabled agents.

    Known limitations

    Requires a standard directory structure for .claude/ or .cursor/. Does not validate the internal logic of external Python or JS scripts.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    Requires project directories to follow standard Claude/Cursor naming conventions. Does not execute or verify external script logic.

    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 buying a skill.

    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 to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed today

    • Free with an Agensi account

    Software
    Python

    Requires Python 3.9+ and a project with a .claude/ or .cursor/ configuration directory. No network or API keys needed.

    Permissions

    Terminal / Shell
    Write Files
    Network Access

    A project with a .claude/ or .cursor/ configuration directory

    Listedtoday

    What's inside

    Frequently Asked Questions

    This skill identifies 'shallow skills'—configs that look like tools but lack executable logic—and fixes 'wrong tool' errors caused by ambiguous MCP descriptions. It also audits your security settings to ensure destructive tools have the necessary confirmation hooks.

    It is purpose-built for developers using Claude Code and Cursor. It specifically scans for .claude/ and .cursor/ directory structures to validate their unique configuration requirements.

    No, this is a local-first Python script that runs entirely on your machine. It never sends your sensitive configuration files, environment variables, or tool definitions to an external server.

    You receive the core Python auditing script, a guide on interpreting the reliability reports, and a library of research-backed patterns for high-performance agent configurations.

    While it checks if your mcp.json contains the necessary environment variables and tool descriptions, it does not execute or debug the internal code of the external servers themselves.

    New in LLM Ops

    Free