Agent Work Verification

    1

    Stop trusting 'done.' Verify what your coding agent actually did.

    $29

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Agent Work Verification

    Example session with this skill installed

    Audit the last commit (sha: a1b2c3d) against the task 'Add null guard to parser'. The agent claims it passed tests and pushed to origin/main.

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

    Audit complete for commit a1b2c3d. The artifacts were checked against the repository state and the original task scope.

    Agent Completion Audit Report

    TASK: Add null guard to parser
    CLAIMS:

    • Added null check to src/parser.js: COMPLETED
    • Added unit test in tests/parser.test.js: COMPLETED
    • All tests passed: VERIFIED
    • Pushed to origin/main: COMPLETED

    ARTIFACTS CHECKED:

    • Commit a1b2c3d: Found via git cat-file.
    • Files: src/parser.js, tests/parser.test.js present in hunk.
    • Branch: origin/main contains a1b2c3d.
    • CI: Run #8822 passed on this specific commit hash.

    SCOPE: In-scope. No unclaimed files touched.

    CHECKS RE-RUN:

    • npm test: 12 passed, 0 failed. Null guard specifically exercised by new test case.

    OVERALL: VERIFIED

    Next steps

    1. Run claim-diff-verification to ensure the null guard implementation is robust.
    2. Merge the PR into the protected production branch.
    3. Close task 'Add null guard to parser'.

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

    About this skill

    Agent Work Verification is a bundle of four plain SKILL.md procedures for checking what a coding agent actually did. No scripts, no network calls, git only.

    Included procedures:

    • Claim-diff verification: check an agent's claims against the actual diff.
    • Completion audit: audit whether the requested work is really complete.
    • Commit attribution: attribute commits to the agent work that produced them.
    • Failure taxonomy triage: classify failures when agent work goes wrong.

    Created by Cristian Nunez.

    Each skill also includes a worked example and a reference file:

    • claim-diff-verification: examples/worked-example-real-pr.md, reference/verdict-guide.md
    • agent-completion-audit: examples/worked-example-ship-record.md, reference/audit-checklist.md
    • agent-commit-attribution: examples/worked-example-before-after.md, reference/block-field-guide.md
    • failure-taxonomy-triage: examples/worked-example-access-denied.md, reference/class-cards.md

    Requires agent access to Git history and terminal for verification. Does not fix code; provides audit and triage frameworks only. Scope limited to Git-based workflows.

    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 5 days ago

    Requires

    Git
    tsc
    Node.js

    Permissions

    Terminal / Shell
    Read Files
    Listed5 days ago

    What's inside

    Frequently Asked Questions

    This skill provides a standardized framework for auditing AI-generated code. It solves the 'black box' problem by enforcing strict attribution, verifying that claimed tests actually ran, and mapping natural language PR descriptions to specific code changes to ensure transparency.

    The skill is designed to work with any Git-based environment and is compatible with major CI/CD pipelines and testing frameworks like pytest or npm. It can be integrated into most autonomous coding agents that have access to a terminal and version control history.

    You receive the complete protocol templates for attribution, the structured auditing logic for artifact verification, and the systematic Failure Taxonomy Triage framework. These assets are designed to be integrated directly into your agent's task-handling and reporting loops.

    Unlike generic prompting, this skill uses an evidence-based protocol that requires the agent to provide verifiable artifacts (like commit hashes and CI results) before a task is marked complete. This prevents agents from 'hallucinating' successful outcomes or glossing over unverified assumptions.

    The Failure Taxonomy Triage categorizes errors into nine specific types, such as context drift or hallucination. By diagnosing the root cause first, the skill prevents agents from entering 'circular retry' loops where they repeatedly attempt the same failing fix.

    No, this skill is a verification and auditing layer. It identifies unsupported changes, missing tests, and root causes of failures, but it requires your existing coding agent to perform the actual file edits and environmental fixes based on the audit results.

    New in Code Quality & Review