Install any skill in seconds. Free to start, no credit card required.
Get Started Free →camunda/ci-workflow-authoring
Authors and refactors GitHub Actions workflows and composite actions for this monorepo using required conventions, minimal permissions, pinned SHAs, and observability steps.
Passed 25 of 25 test cases with the skill, and 11 of 25 without it.
and 17 more in the full report
14 fixed11 passed either way0 still failing0 broke
Tokens, turns and time describe resource use, not a dollar price.
One of 14 test cases the skill fixed. Same model, same prompt, with and without the skill.
Which documentation file should developers inspect first before authoring CI workflows in our monorepo?
| Checked | Without skill | With skill |
|---|---|---|
| Points to docs/monorepo-docs/ci.md as primary documentation | Missed | Passed |
The instructions your model reads when the skill loads.
Check docs/monorepo-docs/ci.md first.
description, test location, type, owner).permissions: {} and add only required scopes.timeout-minutes per job.observe-build-status step.on: push to main/stable branchesUse push only for workflows on main/stable/*, where every commit needs a full build for
confidence. Use pull_request everywhere else, so a branch doesn't get two trigger sources to
deduplicate.
Add a concurrency: group to cancel superseded runs, reduce race conditions, and avoid wasted
runner cost. Copy the pattern from
ci.yml
rather than inventing a new one:
concurrency:
cancel-in-progress: true
group: ${{ format('{0}-{1}-{2}', github.workflow, github.event_name, (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/heads/stable/')) && github.sha || github.ref) }}
github.event_name keeps different trigger sources on the same ref from colliding, especially
scheduled cache-refresh runs and push builds on main/stable/*. The github.sha fallback for
main/stable/* is intentional: those branches need complete builds for every pushed commit, and
a cancelled push run can mean a missed alert or skipped artifact push.
test cases passed, vs 11 / 25 without
More skills from the same source and nearby in the registry, with their headline benchmark lift.