agentleFS
ChariereFiedlerChariereFiedler/ronce-racine/skills/ci-pipeline-orchestration/SKILL.md

ci-pipeline-orchestration

Skill9 starsChanged 25 days agoMITView on GitHub

Use when a CI pipeline needs checking after a push, when CI is red and the failure must be diagnosed, when jobs need retrying, or when validating post-deploy HTTP headers and smoke tests. Triggers on "check the pipeline", "vérifier le pipeline", "rerun CI", "relancer CI", "the pipeline failed", "le pipeline a échoué", "validate the headers", "valider les headers", after any `git push`.

  • Commits and pushes

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 4 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈4

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    This skill: when: a push happened and the CI pipeline must be checked, diagnosed, retried, or its post-deploy validated - CI red, job …

    line 16
  2. Very specific:

    commit-readiness-review: instead if: the errors are local and the code is not pushed yet (pre-commit checks)

    line 17
  3. Very specific:

    production-incident-diagnostic: instead if: production is down or misbehaving outside of an in-progress deploy - that is an incident, not …

    line 18
  4. Distinctive:

    Which commit / branch: should carry the target pipeline (the pushed SHA, not an old one)

    line 31
  5. Distinctive:

    Read an existing neighbor/project skill and copy its commands before inventing

    line 34
  6. Distinctive:

    Smoke test: call the healthcheck / key endpoint, verify a real 2xx.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “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

CI Pipeline Orchestration - drive the pipeline end-to-end after a push

If the current repo has a project-specific pipeline skill, it wins - it knows the CI provider, its commands and the deploy URLs.

This skill vs. others

  • This skill when: a push happened and the CI pipeline must be checked, diagnosed, retried, or its post-deploy validated - CI red, job reruns, HTTP header checks
  • commit-readiness-review instead if: the errors are local and the code is not pushed yet (pre-commit checks)
  • production-incident-diagnostic instead if: production is down or misbehaving outside of an in-progress deploy - that is an incident, not the pipeline

Quick rule: red pipeline after a push → this skill; local errors before push → commit-readiness-review; prod broken outside a deploy → incident diagnostic.

Principle

  • End-to-end: from pipeline status through post-deploy validation (HTTP headers, smoke test) - a green pipeline does not prove the app responds.
  • Diagnose before rerunning: map the failure to a precise cause. Blind reruns mask a real bug and burn runner minutes.
  • Track the recurring: the same failure 2+ times = a ticket, not yet another rerun.

Context to gather (before acting)

  • Which CI provider and how to query it: GitLab CI, GitHub Actions, Jenkins, CircleCI… available CLI (gh run, glab ci, API), or config (.gitlab-ci.yml, .github/workflows/, Jenkinsfile)
  • Which commit / branch should carry the target pipeline (the pushed SHA, not an old one)
  • CI access: token/credentials available to read job status and logs
  • Post-deploy URL (staging/prod) and healthcheck endpoint for validation
  • Read an existing neighbor/project skill and copy its commands before inventing

Protocol

- [ ] 1. Pipeline status of the target commit
- [ ] 2. If failed → diagnose (failed job logs) → map to a cause
- [ ] 3. Fix (push) OR targeted rerun if a genuine flake
- [ ] 4. Post-deploy: smoke test + HTTP headers
- [ ] 5. Recurring failure (2+) → ticket
  1. Current status - list the latest pipelines/runs and fetch the job statuses for the target commit. Identify what is passed, failed, running, manual.
  2. Diagnosis - for each failed job, read the tail of the logs and map it to a cause, not just a symptom:

    Symptom in the logsLikely cause → action
    lint/format diffreformat + fix locally, repush
    red testanalyze the failing test (regression vs flake) before any retry
    build/compile errorreproduce the build locally, fix, repush
    runner infra error / network timeout / image pullinfra flake → legitimate rerun with no code fix
    missing secret/variableCI config to fix, not the code
  3. Act:
    • Real bug: fix on a branch, repush - the pipeline re-triggers. No rerun without a fix.
    • Confirmed infra flake (runner system failure, network timeout): rerun the targeted job. Document the rerun.
    • Manual deploy job: ask the user to confirm before triggering it, especially toward production.
    • Never --no-verify on push even if a hook fails.
  4. Post-deploy - a green pipeline is not enough:
    • Smoke test: call the healthcheck / key endpoint, verify a real 2xx.
    • HTTP headers (curl -I <url>/health):
      • Strict-Transport-Security present in prod
      • X-Content-Type-Options: nosniff
      • Cache-Control consistent (no aggressive caching on an API)
      • Access-Control-Allow-Origin = expected domain
    • App not responding despite a green pipeline → read the deploy logs before declaring success.
  5. Recurrence - the same failure 2+ times → create a CI bug ticket (logs + context + criterion "the job passes on the default branch") instead of rerunning an Nth time.

Templates

Pipeline <id> (<branch>, <sha>): <status>
- Job <a>: ✅/❌ (+ cause if ❌)
- Job <b>: ✅/❌ (+ cause if ❌)
- Smoke test: ✅/❌
- Headers: ✅/❌ (+ detail if KO)
- Action: <fix pushed | flake rerun documented | deploy validated | ticket created>

Traps & rationalizations

ExcuseReality
"I'll rerun, maybe it passes"Rerunning without diagnosing masks a real bug and burns runner minutes. Diagnose first
"It's probably a flake""Flake" is not a decree: a genuine flake = an identified infra error (runner, network). A reproducible red test is not a flake
"This job fails often, I ignore it"An ignored recurring failure = debt that will make the pipeline useless. 2+ times → ticket
"Pipeline green, it's deployed"A green job does not prove the app responds. Smoke test + headers mandatory
"The hook blocks the push, --no-verify"The hook protects the branch. No --no-verify without an explicit request
"I'll trigger the manual deploy directly"Manual job = user confirmation, especially toward prod

Exit condition

  • Target commit's pipeline passed or running (no longer failed)
  • If a fix: commit pushed and a new pipeline triggered
  • If a flake: rerun launched and result documented
  • Post-deploy validated: smoke test 2xx + conformant headers (output pasted, not assumed)
  • If a recurring failure (2+): ticket created

Pipeline still failed with no ticket and no fix = not done.

Tooling

  • The CI provider's CLI depending on the project (gh run view/rerun, glab ci, REST API) - spot the available command before acting, do not hardcode a single provider.

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.