CI Validation Skill

SkillFiles & storage

Lets your agent lint, policy-check, format, and test GitHub Actions workflow changes before you commit them.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the CI Validation Skill skill

About this capability

Validates GitHub Actions workflow changes in this monorepo using actionlint, conftest policy checks, spotless formatting, and act testability assessment. Use when creating, editing, or reviewing workflow files.

What this skill tells your AI

The instructions your AI receives, as published by camunda/camunda in .claude/skills/ci-validation/SKILL.md and read by ahel’s review.

Use this skill for every workflow modification.

Mandatory Sequence

  1. Lint first:
    • actionlint .github/workflows/*.yml
  2. Policy check:
    • conftest test --rego-version v0 -o github --policy .github .github/workflows/*.yml
  3. Format:
    • ./mvnw spotless:apply -T1C
  4. Tier classification (mandatory, before act):
    • For each changed workflow, assign Tier 1/2/3 using .claude/skills/act-testing/references/workflow-tiers.md and the rubric in .claude/skills/act-testing/SKILL.md.
    • Record: workflow path, tier, one-line rationale, and validation path.
  5. Act test (harness-style, mandatory):
    • Read .claude/skills/act-testing/SKILL.md before proceeding.
    • Tier 1 or 2: create a temporary test-*.yml harness, run check-drift.sh, assess act feasibility (act, docker, daemon), and provide user-run scenario commands.
    • Keep test harness files available while sharing commands so users can execute them.
    • Remove test-*.yml files only after user-run act validation is complete.
    • Tier 3: skip harness and document non-applicability with explicit rationale.
    • If multiple workflows changed, either test all Tier 1/2 workflows or explicitly justify why each one is Tier 3.
    • Never run act directly against production workflows.
  6. Iterate until checks are clean.

Do not skip step 1.

Required Evidence Format

For every changed workflow, include:

  • workflow: <path>
  • tier: <1|2|3>
  • reason: <single sentence>
  • evidence: <drift-check + feasibility + scenario matrix | tier-3 rationale>

For Tier 1/2 workflows, evidence must explicitly contain:

  • test-workflow: <path>
  • drift-check: <command + result>
  • feasibility: act/docker/daemon
  • scenario-matrix: <positive + negative at minimum>
  • act-commands: <copy-paste ready plain text using canonical format: act <event> -e <event-json> -W <test-workflow> --secret-file .secrets --reuse; full workspace-relative paths; no markdown links in command text>
  • cleanup: <test harness removed or justification>

If these fields are missing, validation is incomplete and should not be reported as done.

Validation Matrix

  • Simple metadata/security updates: actionlint + conftest + spotless
  • Logic-heavy workflows: actionlint + conftest + spotless + act
  • External-only workflows: actionlint + conftest + spotless + rationale

Preferred VS Code Tasks

  • CI: actionlint (workflows)
  • CI: conftest (workflows policy)
  • CI: verify skill markdown stability

Signals

GitHub stars
4k
Forks
818
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
ci-validation
Source
github.com/camunda/camunda