More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Cuelight Status: Claude Code Agent Status and Map

    1

    Ask "what are my agents doing?" and see which Claude Code subagents failed, stalled, are looping or left code untested, plus an agent map of who started whom and whose work fed which agent. Local, read-only, free.

    Free

    Security scanned
    Cuelight Status: Claude Code Agent Status and Map

    Cuelight Status: Claude Code Agent Status and Map

    Example session with this skill installed

    What are my agents doing? Did anything fail or get stuck? Show me how they worked together.

    • Read your context and instructions
    • Compiled the cuelight status

    Summary

    1 agent failed, 1 went quiet and 1 is repeating the same call. Two agents edited code without testing it, and two agents both wrote types.ts.

    Status

    Checkout rewrite with payments: 13 agents, 8 done, 3 running, 1 failed, 1 stalled
    
    Failed:
      - Write the unit tests (1m 20s)
    
    Stalled (gone quiet):
      - Review the checkout for security issues (41s; last: Bash npm audit --omit=dev)
    
    Running:
      - Build the checkout UI (6m 23s; last: Edit /home/dev/northwind-shop/src/ui/Checkout.tsx)
      - Run the end-to-end suite (2m 20s; repeating npm run e2e -- checkout.spec.ts (possible loop); last: Bash npm run e2e -- checkout.spec.ts)
      - Update the developer docs (27s; last: Write /home/dev/northwind-shop/docs/checkout.md)
    
    Unverified work: 1 failing check(s), 2 agent(s) edited code without a test afterwards.
    Files written by more than one agent: types.ts
    Spent so far: $6.50
    

    Agent map

    I also drew the agent map (who started whom, and whose work fed which agent). Open it in your browser:

    cuelight-reports/agent-map-20261101-1716-demo-che.html
    

    Next steps

    1. Look at why "Write the unit tests" failed (its last npm test failed).
    2. Check on "Review the checkout for security issues": it has been quiet since npm audit.
    3. Stop the end-to-end loop, or fix what it keeps retrying.

    Sample run on a made-up demo session ("Checkout rewrite with payments", 13 subagents).

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

    About this skill

    What it does

    Ask Claude "what are my agents doing?" and get a straight answer about every Claude Code subagent in your session, most urgent first. Ask "show me my agents" for a map of how they worked together.

    What you get

    • Agents that failed, stalled (went quiet) or are repeating the same call.
    • Work never tested after an edit, and files written by two agents.
    • What the session has spent so far (with your own price file).
    • The agent map: every agent in the waves they were launched in, who started whom, and whose work fed which agent. Click an agent to see what it did; hover to light up its connections. Light and dark, works offline.

    How it works

    Claude Code already writes a record of every session and subagent. The skill reads those records on your machine; the map is one offline HTML page.

    Best for

    Anyone running several Claude Code subagents who wants to know which one needs attention.

    Not for

    Alerts the moment an agent waits on your permission (the free Cuelight dashboard plugin does that), or agents other than Claude Code (not yet).

    Want each agent's problems and cost marked on the map, a run score, what to fix next, a budget guard and a check before every pull request? See Cuelight Pro.

    Privacy

    Runs on your machine. Reads only Claude Code's own session records, read-only. Makes no network calls. Credential-shaped text is redacted from everything it writes. Free and open source (MIT).

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.

    • Works with Claude Code only for now.
    • A snapshot when you ask, not a live feed; ask again for an update.
    • Cannot tell when an agent is waiting on your permission (the free Cuelight dashboard plugin can).
    • Spend shows in money only with your own price file; otherwise in tokens.

    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 14 days ago

    • Free with an Agensi account

    Software
    Python 3.9+Claude Code

    Permissions

    Read Files
    Write Files
    Environment Variables
    Terminal / Shell

    File Scopes

    ~/.claude/projects/**
    (read-only:
    Claude
    Code's
    session
    records)
    ~/.config/cuelight/prices.json
    (optional
    read-only:
    your
    own
    prices)
    ./cuelight-reports/agent-map-*.html
    (written
    only
    when
    you
    ask
    for
    the
    map)

    Reads Claude Code's session records read-only. Writes nothing for a status check; writes one self-contained HTML page when you ask for the agent map. No network calls. Environment variables are read only to find folders and settings (where Claude Code keeps sessions, the optional price file, optional tuning); each is listed in SKILL.md. No credentials or tokens are read.

    Listed15 days ago
    Updated14 days ago

    What's inside

    Frequently Asked Questions

    No. It runs on your machine, reads Claude Code's own session records read-only and makes no network calls.

    Not yet. It reads the session records Claude Code keeps, so it works with Claude Code today. Other agents are planned.

    No. The skill works on its own; it only needs Python 3.9 or newer.

    Prices differ by plan, so none are built in. Give Claude your prices once and it saves them to ~/.config/cuelight/prices.json; from then on cost shows in money.

    Cuelight Pro marks each agent's problems and cost on the map, scores the run, ranks what to fix next, and adds Cost Guard and PR Gate.

    New in Agents & Orchestration

    Free