More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    CI Babysitter: Fix Failing PR Checks

    1

    Guide your coding agent through CI logs, local tests, small code fixes, and fresh checks on your pull request.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    CI Babysitter: Fix Failing PR Checks

    CI Babysitter: Fix Failing PR Checks

    Example session with this skill installed

    Use ci-babysitter to fix CI on PR #1. Read the logs, reproduce the failure, fix the source, and push to this PR branch. Keep the tests and workflow unchanged. Stop after three rounds.

    • Read your context and instructions
    • Compiled the ci babysitter

    Real validation session: synthetic GitHub fixture

    CI status: Both existing checks passed after one fix round. Limit: three rounds.
    PR: #1 in our private test repository.

    Before

    Both Node 18 and Node 22 jobs failed. Two assertions expected 90 but received 100. The local command reproduced the same failure: five tests passed and two failed.

    Fix

    The function compared the discount code exactly. The agent added a string check, removed surrounding whitespace, and converted the code to uppercase before comparison. Only validation/ci-babysitter/cart.mjs changed. Tests and the workflow stayed unchanged.

    Baseline commit: f6b8f6e
    Fix commit: 1ab5403
    Failed GitHub run: 38019260674
    Passed GitHub run: 38019416124

    After

    The command node --test validation/ci-babysitter/cart.test.mjs passed all seven tests with zero skips. A normal branch push triggered a fresh CI run. Both Node jobs passed.

    Re-runs: none
    Possibly flaky: none
    Not fixed: none in this fixture
    Next step: Review the code before a merge. The evaluation PR remains a draft.

    No required checks were reported. These results cover both existing checks, not branch-protection requirements. This is one synthetic GitHub case, not a general reliability benchmark. GitLab was not tested.

    Before and after evidence

    Before: both jobs failed

    Before: failed CI

    After: both jobs passed

    After: passed CI

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

    About this skill

    Fix a red pull request

    CI Babysitter gives your coding agent a repeatable process to diagnose failed checks and verify a code fix. Give it a GitHub PR or GitLab MR.

    What you get

    One SKILL.md with CLI commands, failure categories, stop rules, and a report template. This version contains no helper scripts.

    The workflow

    1. Check CLI authentication and the PR branch.
    2. Watch checks and read the failed logs.
    3. Reproduce the failure with the project command.
    4. Make a small fix, run local tests, and push to the PR branch.
    5. Check the new CI run and report the result.

    The default limit is three fix rounds. The agent stops for missing secrets, access problems, or a product decision. The instructions prohibit skipped tests, hidden errors, credential disclosure, and unauthorized force pushes. These rules depend on the agent; they are not software enforcement.

    Verified example

    In our synthetic GitHub test, an independent agent corrected one source line. Both Node 18 and Node 22 jobs then passed. All seven local tests passed. The tests and workflow stayed unchanged. See the demo and before-and-after screenshots.

    Setup and limits

    Use an agent with terminal and file access, Git, an authenticated gh or glab CLI, and the project tools. Installation does not start a background service. GitLab and other failure types were not tested in this evaluation. Review each code change before a merge.

    Instruction-only skill; no standalone background monitor or bundled helper scripts. Requires authenticated gh or glab, Git, project tools, and branch push rights. Default limit: three fix rounds. Cannot fix missing secrets, access rights, billing, or runner outages. Safety instructions depend on the coding agent. One synthetic GitHub case passed; GitLab, flaky retries, and other failure types remain untested. Review code before a merge.

    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

    GitHub
    Gitlab

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access

    Allowed Hosts

    api.github.com
    gitlab.com

    File Scopes

    **/*

    Runs git, gh or glab, and project test commands in the selected repository. Reads source, tests, workflow files, and CI logs. Writes only files needed for the requested fix. Normal pushes target the PR branch. GitHub or GitLab access uses your existing CLI session. Do not supply credentials to the creator. No credential retrieval or unrelated network calls are requested.

    Listed1 month ago

    What's inside

    Frequently Asked Questions

    One SKILL.md with a CI procedure, CLI commands, failure categories, stop rules, and a report template. This version contains no helper scripts or background service.

    Use an agent with terminal and file access. Install Git and gh or glab. The skill checks CLI authentication first. If authentication fails, it stops and asks you to log in. Provide the project tools and permission to push to the PR branch.

    The workflow uses normal commits and pushes to the requested PR branch. Its rules prohibit skipped tests and weaker assertions. It asks before a force push or a CI setting change. Review the resulting code; these instructions depend on the agent.

    An independent agent used the skill on a synthetic GitHub PR. Two Node jobs failed initially, then passed after one source fix. All seven local tests passed. The tests and workflow stayed unchanged. See the demo and screenshots. GitLab was not tested.

    The default limit is three fix rounds. It stops if the same failure returns after a fix, or if it needs secrets, access rights, CI settings, or a product decision. It may retry a suspected flaky or infrastructure job once. A second retry needs user approval.

    New in DevOps & Cloud