More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Verify Agent Work: independent QA reviewer

    1

    Stop trusting an agent's own summary. A fresh-context reviewer re-checks every claim against files, logs and a re-run, then writes a PASS / FIX / FAIL verdict with evidence.

    $9

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Verify Agent Work: independent QA reviewer

    Example session with this skill installed

    Review the work done by the code-agent in /src/auth. I have the report in agent-log.txt. Check it against the original ticket to ensure all tests pass and no secrets were committed.

    • Read your context and instructions
    • Compiled the verify agent work

    I have completed the independent review of the code-agent work in /src/auth. I compared the claims in agent-log.txt against the actual filesystem state and re-ran the authentication test suite.

    QA-auth-module-rev1.md

    Verdict: FIX
    Checked: 5 claims, 4 confirmed, 1 wrong, 0 could not check
    Re-run: npm test src/auth/auth.test.ts → 12 passing, 1 failing (Auth-04: Token Expiry)
    
    Findings:
    1. "All tests pass" — WRONG — `auth.test.ts:42` shows a failure in token expiration logic.
    2. "No secrets committed" — PASS — `grep` scan of `.env` and `src/auth/` returned no matches.
    3. "Added login endpoint" — PASS — Found `POST /api/login` implementation in `auth.controller.ts:15`.
    4. "Updated README" — PASS — Documentation reflects new auth flow.
    5. "Removed debug logs" — PASS — Verified `console.log` removal in production paths.
    
    Fix list (FIX only):
    1. Fix the failing test case Auth-04 in `src/auth/auth.service.ts` where token TTL is not being applied correctly.
    2. Align the internal error message in `auth.controller.ts` with the brief's requirement for generic error responses.
    

    Next steps

    1. Review the failing test in src/auth/auth.test.ts to identify the logic gap.
    2. Send the fix list back to the code-agent for correction.
    3. Re-run this verification skill once the FIX verdict items are addressed.
    QA-auth-module-rev1.md

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

    About this skill

    Coding agents often report work as finished when a test still fails or a step was skipped. This skill adds an independent QA pass: a reviewer with a fresh context treats the builder's report as a list of claims, not facts, and checks each one against the files, logs and a re-run.

    What it does

    • Brief check: compares the deliverable with what was asked, including anything done outside scope.
    • Claim check: lists every factual claim in the report (files, counts, "tests pass", "committed") and checks each one.
    • Re-run: runs at least one thing end to end (tests, build, the script that produced a number) and records the actual output line.
    • Bright lines: confirms nothing was sent, published, pushed, bought or exposed.
    • Verdict file: writes QA-<item>.md with PASS, FIX or FAIL, the evidence, and a full fix list.

    How it works

    1. Give your agent the brief, the builder's report and the deliverable paths.
    2. Ask it to "verify the result of <task>".
    3. The reviewer works through the included verification-checklist.md and writes the verdict next to the deliverable.

    What's included

    • SKILL.md: the reviewer instructions
    • verification-checklist.md: cheap checks that catch the most common false claims
    • README.md: install notes

    Known limitations

    The reviewer reports; it does not fix. That is on purpose: a reviewer that fixes quietly hides the gap. It needs read access to the files and permission to run your tests or build.

    Independent work by Lerider Technologies. Not affiliated with or endorsed by Anthropic.

    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.

    Reports only; it does not fix what it finds (on purpose). Needs read access to the project and permission to run your tests or build.

    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

    Money-back guarantee

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

    Trust & safety

    Security scan

    Security scan passed 5 days ago

    Software
    Git

    Permissions

    No special permissions declared or detected

    Listed5 days ago
    Updated3 days ago

    What's inside

    Frequently Asked Questions

    An agent checking its own work tends to agree with itself. This skill runs the check as a separate reviewer with a fresh context, treats the builder's report as claims to test, and requires evidence (a file, a log line, a re-run output) for every PASS.

    Any agent that loads the SKILL.md format, such as Claude Code, Codex CLI, Cursor or Gemini CLI. It works best where the reviewer can be started as a separate sub-agent or a new session.

    No. It writes a PASS / FIX / FAIL verdict with a full fix list and leaves the fixing to the builder. A reviewer that fixes quietly hides the gap.

    Nothing beyond your agent: unzip the folder into your skills directory. The reviewer needs read access to the project and permission to run whatever proves the work (tests, build, a script).

    Three files: SKILL.md (the reviewer instructions), verification-checklist.md (cheap checks that catch common false claims) and README.md (install notes).

    A QA-<item>.md file next to the deliverable: the verdict, how many claims were checked, confirmed, wrong or uncheckable, the re-run command with its actual output line, numbered findings, and for FIX the complete fix list.

    New in Agents & Orchestration