github-actions— agent skill

You are in AUTONOMOUS MODE. Do NOT ask questions. Do NOT pause for confirmation. Execute every phase below in sequence, making decisions based on what you find.

============================================================ PHASE 0 — INPUT ============================================================

$ARGUMENTS may contain:

• A deploy target: vercel, aws, gcp, fly, cloudflare, docker-registry, npm, pypi
• A scope: ci-only (just PR checks), cd-only (just deploy), full (both — default)
• --preview — include preview deployment workflow for PRs
• --matrix — include matrix testing across multiple versions
• --monorepo — handle monorepo with path filters
• Specific workflow name to generate (e.g., test, deploy, release)

============================================================ PHASE 1 — STACK DETECTION ============================================================

Detect the project stack by scanning for manifest files:

Node.js / TypeScript:

• package.json — read scripts (test, build, lint, typecheck, start)
• Lock file: package-lock.json (npm), yarn.lock (yarn), pnpm-lock.yaml (pnpm)
• Framework: check dependencies for next, nuxt, remix, astro, fastify, express, nestjs
• tsconfig.json — TypeScript project

Python:

• requirements.txt, pyproject.toml, setup.py, Pipfile
• Framework: django, flask, fastapi, starlette
• Tool: poetry, pip, pipenv, uv

Go:

• go.mod — read module name and Go version
• Check for cmd/ directory structure

Rust:

• Cargo.toml — read workspace members if monorepo

Flutter / Dart:

• pubspec.yaml — read dependencies, check for flutter_test

Docker:

• Dockerfile — if exists, prefer Docker-based CI steps

Record: language, package manager, framework, test command, build command, lint command.

============================================================ PHASE 2 — GENERATE PR CHECK WORKFLOW ============================================================

Create .github/workflows/ci.yml:

name: CI
on:
  pull_request:
    branches: [main, master]
  push:
    branches: [main, master]

Include these jobs in order:

1. Lint & Format

• Node: npm run lint / eslint . / prettier --check .
• Python: ruff check . / black --check . / mypy .
• Go: golangci-lint run
• Rust: cargo clippy -- -D warnings / cargo fmt --check

2. Type Check (if applicable)

• Node/TS: tsc --noEmit or npm run typecheck
• Python: mypy . or pyright

3. Test

• Node: npm test with coverage (--coverage)
• Python: pytest --cov or python -m pytest
• Go: go test ./... -race -coverprofile=coverage.out
• Rust: cargo test
• Flutter: flutter test --coverage

4. Build

• Node: npm run build
• Go: go build ./...
• Rust: cargo build --release
• Flutter: flutter build apk --debug (or web)
• Docker: docker build . (if Dockerfile exists)

5. Security Scan

• Always include: actions/dependency-review-action@v4 on PRs
• Node: npm audit --audit-level=high
• Python: pip-audit or safety check
• Go: govulncheck ./...
• Rust: cargo audit

Caching — always include appropriate caching:

• Node: actions/setup-node@v4 with cache: 'npm' (or yarn/pnpm)
• Python: actions/setup-python@v5 with cache: 'pip'
• Go: actions/setup-go@v5 with cache: true
• Rust: Swatinem/rust-cache@v2
• Flutter: subosito/flutter-action@v2 with cache: true

Matrix testing (if --matrix):

• Node: test against LTS and current (e.g., [18, 20, 22])
• Python: [3.10, 3.11, 3.12]
• Go: last two minor versions
• OS matrix: [ubuntu-latest] default, add macos-latest if Flutter/mobile

Permissions: Use minimal permissions per job (never write-all).

============================================================ PHASE 3 — GENERATE DEPLOY WORKFLOW ============================================================

Create .github/workflows/deploy.yml based on deploy target:

Vercel: Use amondnet/vercel-action@v25 or Vercel CLI. Production deploy on push to main. Preview deploy on PR (if --preview).

AWS (ECS/ECR): Configure AWS credentials via OIDC (aws-actions/configure-aws-credentials@v4), login to ECR, build/tag/push Docker image, update ECS service.

AWS (S3 + CloudFront): Sync build output to S3, invalidate CloudFront distribution.

GCP (Cloud Run): Auth with google-github-actions/auth@v2, deploy with google-github-actions/deploy-cloudrun@v2.

Fly.io: Setup flyctl, deploy with flyctl deploy --remote-only.

Cloudflare (Pages/Workers): Use cloudflare/wrangler-action@v3.

Docker Registry (GHCR): Login with docker/login-action@v3, build/push with docker/build-push-action@v5 with layer caching. Tag with SHA and latest.

npm / PyPI: Publish on release tag creation. npm: npm publish with NODE_AUTH_TOKEN. PyPI: pypa/gh-action-pypi-publish@release/v1.

Include environment protection rules:

environment:
  name: production
  url: ${{ steps.deploy.outputs.url }}

============================================================ PHASE 4 — GENERATE ADDITIONAL WORKFLOWS ============================================================

If applicable, also create:

Release workflow (.github/workflows/release.yml):

• Triggered on tag push (v*)
• Create GitHub Release with changelog
• Build and attach artifacts

Preview deploy (if --preview):

• Comment on PR with preview URL
• Clean up preview on PR close

Dependabot (.github/dependabot.yml):

• Always create if it does not exist
• Include github-actions ecosystem with weekly schedule
• Include the project's package ecosystem (npm, pip, gomod, cargo, etc.)

============================================================ PHASE 5 — SECRETS DOCUMENTATION ============================================================

Create or update .github/SECRETS.md listing all required repository secrets:

## Required Secrets

| Secret | Description | Where to get it |
|--------|-------------|-----------------|
| DEPLOY_TOKEN | Vercel/Fly/etc deploy token | {provider dashboard URL} |

List every ${{ secrets.* }} reference used in the generated workflows.

============================================================ SELF-HEALING VALIDATION (max 2 iterations) ============================================================

After completing deployment/infrastructure changes, validate:

1. Verify all generated files are syntactically valid (YAML, JSON, HCL, Dockerfile).
2. Run validation commands if available (terraform validate, docker build --check, kubectl dry-run).
3. Verify no secrets, credentials, or sensitive values are hardcoded.
4. If validation fails, diagnose and fix the specific syntax or config error.
5. Repeat up to 2 iterations.

IF STILL FAILING after 2 iterations:

• Document what failed and the exact error
• Include partial output if available

============================================================ OUTPUT ============================================================

Print a summary of all generated files:

## GitHub Actions Setup Complete

### Files Created
- .github/workflows/ci.yml — PR checks (lint, typecheck, test, build, security)
- .github/workflows/deploy.yml — Deploy to {target} on push to main
- .github/dependabot.yml — Automated dependency updates
- .github/SECRETS.md — Required secrets documentation

### Required Secrets
{list of secrets to configure in GitHub repo settings}

### Detected Stack
- Language: {language}
- Framework: {framework}
- Package Manager: {pm}
- Deploy Target: {target}

============================================================ NEXT STEPS ============================================================

1. Add required secrets to GitHub repository settings
2. Push workflows to trigger the first run
3. Enable branch protection rules requiring CI to pass before merge
4. Review caching strategy after first run to verify cache hits
5. Enable GitHub secret scanning and push protection on the repository

============================================================ SELF-EVOLUTION TELEMETRY ============================================================

After producing output, record execution metadata for the /evolve pipeline.

Check if a project memory directory exists:

• Look for the project path in ~/.claude/projects/
• If found, append to skill-telemetry.md in that memory directory

Entry format:

### /github-actions — {{YYYY-MM-DD}}
- Outcome: {{SUCCESS | PARTIAL | FAILED}}
- Self-healed: {{yes — what was healed | no}}
- Iterations used: {{N}} / {{N max}}
- Bottleneck: {{phase that struggled or "none"}}
- Suggestion: {{one-line improvement idea for /evolve, or "none"}}

Only log if the memory directory exists. Skip silently if not found. Keep entries concise — /evolve will parse these for skill improvement signals.

============================================================ DO NOT ============================================================

• Do NOT use deprecated actions (e.g., actions/checkout@v2 — use v4)
• Do NOT hardcode secrets — always use ${{ secrets.NAME }}
• Do NOT use actions/setup-node@v3 — use v4
• Do NOT skip the security scanning step
• Do NOT use ubuntu-20.04 — use ubuntu-latest or ubuntu-24.04
• Do NOT add permissions: write-all — use minimal permissions per job
• Do NOT generate workflows for stacks not detected in the project
• Do NOT overwrite existing workflow files without reading them first and preserving custom steps
• Do NOT use npm ci without checking that package-lock.json exists (use npm install if no lock file)