agentleFS
RemiMyrsetRemiMyrset/roots/.agents/skills/fix-ci/SKILL.md

fix-ci

Skill0 starsChanged 43 days agoMITView on GitHub

Make a red CI run green by reproducing the failing step locally and fixing it at the source. Use when the user says "fix CI", "the checks are red", "why did CI fail", "CI is failing", or after the pr skill reports a red run. Also use unprompted when a PR you opened shows a failed check. Never loosens a checker, never bypasses a hook, never retries a run hoping it passes.

39 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use get-pr-comments, create-pr and 7 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • jaydubya818Skill

    get-pr-comments

    Use when pull-request feedback needs to be fetched and summarized for action.

    Also in 3 projects that use this

  • SentrySkill

    create-pr

    Create pull requests following Sentry conventions. Use when opening PRs, writing PR descriptions, or preparing changes for review. Follows Sentry's code review guidelines.

    Also in 10 projects that use this

  • MicrosoftSkill

    merge

    Merge changes from the topic branch to the merge base branch. Use when the user wants to merge their session's work back to the base branch.

    Also in 3 projects that use this

  • SentrySkill

    commit

    Create commit messages following Sentry conventions. Use when committing code changes, writing commit messages, or formatting git history. Follows conventional commits with Sentry-specific issue references.

    Also in 10 projects that use this

  • NVIDIASkill

    testing

    Run Python and Rust tests for physicsnemo-curator using uv + pytest with coverage reporting via pytest-cov, and cargo-nextest for Rust tests. Includes benchmark workflows with pytest-benchmark and criterion.

    Also in 3 projects that use this

  • AWSSkill

    ship

    Finalize dev work on an issue branch — update docs, push, and open a PR.

    Also in 3 projects that use this

  • +1 more from getsentry

The file

We split this skill into the rules it gives an agent

2 rules. 0 are common, 1 is distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 37 edited versions we read keep these.

  • 1

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Read the first error, not the last line.

    line 15
  2. Very specific:

    If the failing gate is a template mechanic (pnpm verify marks them "template mechanics": test:hooks, test:sync, test:docs, test:gates), …

    line 29

“Common” is measured: kept by at least 6 in 10 of the 37 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Fix CI

CI runs the same steps as pnpm verify, in the same order, on Ubuntu and Windows under bash; scripts/test-gates.mts keeps the two equal. So every red step has a local twin, and the fix is local.

  1. Find the failure. gh pr checks <n> (or gh run list --branch <branch>, the only route for a run on main, which has no PR number) names the red job; gh run view <run-id> --log-failed prints only the failing step's log. Read the first error, not the last line.
  2. Reproduce. For a red run on main, update the local copy first (git switch main, then git pull --ff-only): a stale main reproduces an older commit, and the branch step 5 cuts would start from it. Map the step to its gate and run pnpm verify --only <gate> (pnpm verify <gate> resumes from there through the rest). A step that is red on Windows only usually means a path separator, a CRLF, or a spawn difference; the log says which.
  3. Fix at the source. A failing check means the code, the doc, or the lockfile is wrong, never the checker: do not loosen a rule, skip a test, or add --no-verify (the guards deny it anyway). A stale lockfile is pnpm install, then commit pnpm-lock.yaml. Stale generated docs, or merge conflict lines inside a generated region, are pnpm docs:gen, which rewrites the whole region; commit the output and never hand-merge one.
  4. If the failing gate is a template mechanic (pnpm verify marks them "template mechanics": test:hooks, test:sync, test:docs, test:gates), read which file the failure names. A file this repository owns is a local fix like any other: its own workflow (a pnpm step that is deliberately not a gate ends its line with # not a gate), AGENTS.md (a Commands list that leaves out a gate), .claude/settings.json, or package.json. A failure inside the synced files themselves is not a child's to patch; say so and point at the sync-template skill, or report it upstream.
  5. Prove it. Run pnpm verify end to end. If git branch --show-current matches a protected pattern (a red run on main after a merge), run git switch -c fix/<slug> first: the push guard denies the protected branch. Then commit with a Conventional subject (fix(scope): what, at most 50 chars).
  6. Push. From a new branch, hand it to the pr skill; otherwise push the same branch. Watch the PR with gh pr checks <n> --watch, report the outcome, and stop. In Claude Code the gh calls run without a prompt; Codex and Gemini ask for each.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.