Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Redpen — Verify Coding-Agent Work

    1

    Check coding-agent claims against fresh tests, builds and Git evidence. Return a clear verification receipt.

    Free

    1 useSecurity scanned
    Redpen — Verify Coding-Agent Work

    Redpen — Verify Coding-Agent Work

    Example session with this skill installed

    Use Redpen for this task. Fix null cursor pagination, add a regression test that fails on the original implementation and passes on the fix, and return an evidence receipt. Keep any unsupported compatibility claims visible.

    • Read your context and instructions
    • Compiled the redpen

    Reproducible CLI demonstration

    The receipt below was generated by the published redpen-cli@0.4.0 on a disposable Git repository. The bug fix, regression test and completion claims were scripted fixtures; this is real CLI verification, not a recording of an Agensi agent session. The baseline was saved before the fix.

    Redpen evidence receipt

    Task: Fix null cursor pagination

    Verdict: DONE (1 agent claim remains unverified)
    Checked: 2026-10-17T08:43:51.824Z
    Summary: 9 proven · 0 failed · 1 unverified
    Claim policy: Required criteria must be proven; contradicted claims block DONE

    Definition of done

    • ✓ PROVEN — Implementation changed (required)
      • 2 files changed · +2 -1
      • Evidence: Changes made after the task started were found.
    • ✓ PROVEN — Regression coverage (required)
      • tests/null-cursor.test.mjs
      • Evidence: 1 test file changed.
    • ✓ PROVEN — Tests pass (required)
      • npm run test · exit 0
      • Evidence: npm run test → exit 0 in 185ms
    • ✓ PROVEN — Build passes (required)
      • npm run build · exit 0
      • Evidence: npm run build → exit 0 in 121ms
    • ✓ PROVEN — Null cursor regression reproduces the bug (required)
      • “null cursor starts at beginning” fails with an assertion on the baseline and passes on the current snapshot. This proves the tested case, not general correctness.
      • Evidence: node --test tests/null-cursor.test.mjs · exit 1
      • Evidence: node --test tests/null-cursor.test.mjs · exit 0

    Agent claims

    • ✓ PROVEN — “Updated src/page.mjs.”
      • src/page.mjs · +1 -1
      • Evidence: src/page.mjs changed after this task started.
    • ✓ PROVEN — “Added regression coverage.”
      • tests/null-cursor.test.mjs
      • Evidence: 1 test file changed.
    • ✓ PROVEN — “All tests pass.”
      • npm run test · exit 0
      • Evidence: npm run test → exit 0 in 185ms
    • ✓ PROVEN — “Build succeeds.”
      • npm run build · exit 0
      • Evidence: npm run build → exit 0 in 121ms
    • ? UNVERIFIED — “No breaking changes.”
      • No deterministic evidence is available. Redpen doesn't guess.

    Generated by Redpen. Command output and the agent's full completion are omitted; review the receipt before sharing.

    What happens next

    The required task criteria passed. “No breaking changes” remains unverified; tests do not establish general compatibility. A separate strict-claims run returned NOT DONE. After changing the implementation, the saved receipt returned CHECK NEEDED, and a fresh verification failed. Review unresolved claims before relying on the handoff.

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

    About this skill

    “Done” is a claim. Evidence makes it true.

    Redpen helps your coding agent establish what a task requires, record its completion claims, and check those claims against fresh repository evidence before handing work back.

    What you get

    • A task-specific definition of done established before editing.
    • Independent test and build execution, with Git evidence for implementation and test changes.
    • A readable receipt separating PROVEN, FAILED and UNVERIFIED items.
    • Stale-evidence detection when files change after a check.
    • Optional strict claim verification and a focused JavaScript regression-proof beta.

    Start here

    In a shell-capable coding agent with your local Git repository open, ask: “Use Redpen for this task. Fix the bug, add regression coverage, and return an evidence receipt.”

    This free Agensi edition provides the workflow instructions. The runtime is the separately published, MIT-licensed redpen-cli@0.4.0 package. Node.js 18+, npm and Git are required. First use may download that pinned package from npm after any required permission; no global installation or project dependency changes are needed.

    Honest limits

    Passing tests do not prove general correctness, API compatibility or the absence of breaking changes. Unsupported claims remain UNVERIFIED. The focused regression-proof beta supports one named, dependency-free JavaScript node:test assertion case; richer projects use their normal configured checks.

    This edition has no completion hooks, hosted verification service or automatic transcript capture. Ordinary web chat without shell and repository access cannot run it. Agent-specific session adapters are not promised.

    Local execution and privacy

    Redpen runs locally and has no telemetry or hosted backend. Agensi and your agent provider have their own data policies. Project checks execute repository code with the environment’s permissions and may access files or the network. Receipts can contain project-derived information: review before sharing.

    Built and maintained by Chirag Arora. Source, documentation and issue reporting: https://github.com/ChiragArora31/Redpen

    Instruction-only Agensi package; the runtime is separately retrieved from npm as redpen-cli@0.4.0. Requires Node.js 18+, npm, Git, shell access and a local Git repository. No semantic correctness or general API-compatibility proof. Focused regression proof is a dependency-free JavaScript node:test beta. No completion hooks or automatic transcript capture. CLI behavior has been tested; Agensi-mediated agent behavior has not yet been independently validated.

    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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 1 month ago

    • Free with an Agensi account

    Requires

    Node.js 18+
    npm
    Git
    Terminal access
    redpen-cli 0.4.0
    Node.js
    redpen-cli

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables

    Runs local Git and project test/build commands with the client’s existing permissions. Writes local .redpen task state and reports. Initial pinned npm package retrieval requires network access; project checks may also use network and inherited environment variables. No Redpen service credentials or telemetry. This skill does not install or enable hooks. Review commands and receipts before execution or external sharing.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    Yes. This listing is free and provides instructions for the MIT-licensed Redpen CLI. The executable is separately available as redpen-cli@0.4.0 from npm; it is not bundled into this skill. First use may download the pinned package after any permission required by your agent.

    Verification requires a shell-capable agent with Node.js 18+, npm, Git and access to your local repository. Installing a skill does not grant those capabilities. Without them, the skill explains the missing prerequisites and does not generate a verdict.

    No. DONE means the configured required criteria were proven and no claim was contradicted under the selected policy. Unsupported claims remain UNVERIFIED. Strict claims mode requires every claim to be proven. Passing tests or changed files do not prove general correctness or API compatibility.

    It can run fresh project checks, but it cannot invent a historical baseline or attribute earlier edits to the task. For task-specific evidence, start Redpen before implementation.

    No. This Agensi edition records a concise completion summary via redpen claims and verifies it before returning a receipt. No hooks, automatic transcript capture, new credentials or Redpen account are required.

    Open an issue at https://github.com/ChiragArora31/Redpen/issues with your runtime version, expected behavior and a minimal reproduction. Remove secrets, private source and transcripts. For security issues, follow the repository SECURITY.md.

    New in Code Quality & Review