ci-automation

randlee's avatarfrom randlee

Run CI quality gates with optional auto-fix and PR creation.

⭐0stars🔀0forks📁View on GitHub🕐Updated Dec 21, 2025
[CI and CD]

When & Why to Use This Skill

The CI Automation skill is a comprehensive DevOps tool for Claude that automates the end-to-end development workflow, including pulling code, building, testing, and creating pull requests. By integrating specialized agents for root cause analysis and automated bug fixing, it ensures code quality gates are met while significantly reducing manual intervention in the CI/CD pipeline.

Use Cases

  • Automated Quality Assurance: Automatically execute build and test suites to validate code changes against defined quality gates before merging.
  • Self-Healing Pipelines: Detect straightforward build or test failures and apply automated fixes to resolve issues without developer manual effort.
  • Streamlined PR Workflows: Manage the entire process of pulling target branches, resolving minor conflicts, and submitting pull requests once all checks pass.
  • Intelligent Debugging: Utilize the root-cause agent to summarize complex failures and provide actionable recommendations for unresolved issues in the CI environment.
nameci-automation
version0.7.0
descriptionRun CI quality gates with optional auto-fix and PR creation.
entry_point/ci-automation

CI Automation Skill

Coordinates pull → build → test → fix → PR. Keeps instructions lean; heavy details live in references and contracts file. Agent Runner is required for all agent calls (registry-enforced, audited).

Commands

  • /ci-automation (flags: --build, --test, --dest, --src, --allow-warnings, --yolo, --help)

Agents (v0.1.0)

  • ci-validate-agent: pre-flight checks (clean repo, config present, auth available)
  • ci-pull-agent: pull target branch, handle simple conflicts
  • ci-build-agent: run build, classify failures
  • ci-test-agent: run tests, classify failures/warnings
  • ci-fix-agent: apply straightforward fixes
  • ci-root-cause-agent: summarize unresolved issues, recommend actions
  • ci-pr-agent: commit/push/PR when gates pass

Flow

  1. Validate via Agent Runner → ci-validate-agent
  2. Pull via Agent Runner → ci-pull-agent (dest inferred from tracking; --dest overrides)
  3. Build via Agent Runner → ci-build-agent
  4. On build fail: Agent Runner → ci-fix-agent (only straightforward fixes), repeat
  5. Test via Agent Runner → ci-test-agent
  6. On test/warn fail: Agent Runner → ci-fix-agent if straightforward, else ci-root-cause-agent
  7. If clean and confirmed (or --yolo): Agent Runner → ci-pr-agent to commit/push/PR

Config

Preferred: .claude/ci-automation.yaml (fallback: .claude/config.yaml) with upstream_branch, build_command, test_command, warn_patterns, allow_warnings, auto_fix_enabled, pr_template_path, repo_root. Detect stack (e.g., .NET/Python/Node) and prompt to save suggested commands.

Safety

  • No force-push; respect protected branches.
  • Default is conservative: auto-fix only; stop before commit/PR unless clean and confirmed.
  • --yolo: allow commit/push/PR after gates pass.
  • Warnings block PR unless --allow-warnings or config override.
  • Explicit confirmation for PRs to main/master unless --dest main provided.
  • Audit: Agent Runner logs to .claude/state/logs/ci-automation/.

References

  • .claude/references/ci-automation-commands.md
  • .claude/references/ci-automation-checklists.md
  • .claude/references/ci-automation-contracts.md

Size Guidance

  • Keep SKILL.md under ~2KB: overview, flags, agent list, flow summary, contracts, config, safety, references.