Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    agent context audit

    1

    Audit and optimize local agent folders to fix version drift, deduplicate skills, and reclaim context tokens.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    agent context audit

    agent context audit

    Example session with this skill installed

    Run a scan on my local agent folders and let me know why my Claude Code context is so high.

    • Read your context and instructions
    • Compiled the agent context audit
    • Generated the document

    I have completed the scan of your local agent directories. It looks like you are carrying significant overhead from duplicate skills and a few oversized descriptions in your Claude Code profile.

    Agent Context Audit Summary

    AgentTotal SkillsUniqueAlways-on TokensWasted (Dupes)MCP Servers
    Claude Code1491,8406203
    Cursor8891001
    Gemini CLI544401100

    Top Findings

    Duplicate (duplicate): git-helper is installed in both Claude Code and Gemini CLI folders. Identical content hashes.
    2.

    Version Drift (version-drift): deploy-script has different content in Claude Code vs Cursor. The Claude Code version is 4 days newer.
    3. Oversized Description (large): documentation-expert has a SKILL.md body over 650 lines.
    4.

    Invalid Metadata (invalid): test-runner is missing a description field, causing the agent to potentially ignore it.
    5. Context Waste: You are losing ~620 tokens at every Claude Code startup due to redundant plugins.

    The full interactive report is available at: ./audit_reports/report_2023-10-27.html

    Next steps

    1. Run claude plugin disable git-helper to remove the redundant CLI instance.
    2. Review the diff for deploy-script and I will symlink the latest version to Cursor.
    3. Move the long-form examples in documentation-expert to a references/ folder to save 400+ tokens.

    report_2023-10-27.html

    PDF · document

    Generated

    Example file from a real run - the skill writes it into your workspace.

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

    About this skill

    Local agent environments often become cluttered with redundant skills, outdated plugins, and conflicting instructions that drain your context window. This skill audits your local machine to identify why your agents are feeling slow or hitting token limits before you even start coding. It scans the configuration folders for Claude Code, Cursor, Codex, Gemini CLI, and OpenClaw to find duplicate skills, version drift, and oversized metadata files.

    What it does

    • Deduplicate skills identifies identical skill files stored in multiple agent directories to reclaim context space.
    • Detect version drift finds skills with the same name but different content, preventing inconsistent agent behavior.
    • Analyze token overhead provides estimates of "always-on" token costs per agent based on loaded skill descriptions.
    • Identify description overlap flags skills with similar descriptions that confuse agents into picking the wrong tool.
    • Validate metadata checks for missing names, broken symlinks, or descriptions exceeding 1024 characters.

    How it works

    1. Scan run the audit script to crawl agent folders and generate an audit.json and an HTML report.
    2. Review findings analyze the summary table of always-on tokens, wasted context, and specific errors like version-drift or large.
    3. Plan cleanup review a proposed list of fixes including disabling plugins, sharpening descriptions, or merging files.
    4. Execute & Verify apply the approved changes and re-run the scan to confirm the new, lower token baseline.

    Frameworks & tools

    Built for Node.js 18+. Compatible with Claude Code, Cursor, Codex, Gemini CLI, and OpenClaw on Windows, macOS, and Linux.

    Why this beats prompting it yourself

    Standard agents cannot see their own installation directories or the global config folders of competing agents. This skill provides the file-system access and cross-agent analysis required to see the "hidden" context costs you pay at every startup.

    Use cases

    • Debugging why an agent frequently hits context limits in small projects.
    • Syncing custom skills across different IDEs and CLI tools.
    • Auditing local MCP server overhead and skill description quality.
    • Cleaning up legacy skill versions after tool updates.

    Known limitations

    Token counts are estimates based on character length. MCP tool counts are restricted to server counts as live connections are not established during the scan.

    Token counts are estimates. MCP tool counts reflect server count only, not individual tools. Detection relies on known agent folder locations which may change between agent versions.

    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

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    Requires

    macOS
    Node.js

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    This skill identifies hidden inefficiencies like duplicate skill files, version drift, and bloated descriptions that drain your context window before you even start a chat. By cleaning these up, you reclaim tokens and ensure your agent selects the correct tools without confusion.

    It is fully compatible with Claude Code, Cursor, Codex, Gemini CLI, and OpenClaw across Windows, macOS, and Linux. The skill scans the specific configuration and local storage directories these tools use to store their instructions and skills.

    No, the skill performs a local file-system scan and generates a report for your review first. You maintain full control, as the skill provides a proposed list of fixes that you can approve and verify before any changes are applied to your configuration.

    The audit requires Node.js 18 or higher installed on your machine. Once installed, you simply point the script at your agent configuration folders to generate a detailed HTML report and an audit.json file.

    Token usage is calculated based on character-to-token estimates for skill descriptions and system prompts that are loaded by default. This gives you a clear baseline of the 'always-on' context cost your agent incurs at startup.

    Your purchase includes the full audit toolset, the reporting engine for HTML/JSON summaries, and a guide for interpreting 'version drift' and 'description overlap' errors.

    New in LLM Ops