asepharyana-hub-guide/skills/ci-cd/SKILL.md

---
name: ci-cd
description: CI/CD best practices — GitHub Actions, pipeline design, Docker build + push, deployment workflows, testing in CI, and environment management. Use when designing CI/CD pipelines, debugging workflow failures. Triggers from project files and configuration, not just keyword matching."
---

# CI/CD Best Practices

## Pipeline Design Principles

1. **Fast feedback** — failing fast is better than failing comprehensively.
2. **Deterministic** — same commit = same result, same artifacts.
3. **Immutable artifacts** — build once, promote through environments.
4. **Idempotent deployments** — deploying the same artifact again produces the same result.
5. **Security gates** — scan dependencies, secrets, and code before production.

## GitHub Actions Structure

```
.github/workflows/
├── lint.yml                     # Quick checks — runs in <2 min
├── docker-build-push.yml        # Build + push images
├── deploy-docker.yml            # Deploy to VPS
├── security.yml                 # CodeQL + dependency scan
├── update-submodule.yml         # Update submodule pointer
└── ...other workflows
```

### Workflow Patterns

**Lint (fast — gates everything else)**
```yaml
name: Lint
on: [push, pull_request]
jobs:
  lint:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: oven-sh/setup-bun@v2
      - run: bun install --frozen-lockfile
      - run: bun run ci     # Biome lint + format
```

**Build + Push (on push to main)**
```yaml
name: Build Docker
on:
  push:
    branches: [main]
    paths: ['apps/**', 'infra/**']
  repository_dispatch:
    types: [build]

jobs:
  build:
    strategy:
      matrix:
        service: [frontend, api]
    steps:
      - uses: actions/checkout@v4
      - run: |
          docker build \
            -f infra/docker/${{ matrix.service }}.Dockerfile \
            -t ghcr.io/myorg/myproject/${{ matrix.service }}:sha-${{ github.sha }} \
            -t ghcr.io/myorg/myproject/${{ matrix.service }}:latest \
            .
      - run: docker push --all-tags ghcr.io/myorg/myproject/${{ matrix.service }}
```

**Deploy (after build)**
```yaml
name: Deploy
on:
  workflow_run:
    workflows: ['Build Docker']
    types: [completed]
  push:
    branches: [main]
    paths: ['infra/**']

jobs:
  deploy:
    runs-on: ubuntu-latest
    steps:
      - uses: appleboy/ssh-action@v1
        with:
          host: ${{ secrets.VPS_HOST }}
          username: ${{ secrets.VPS_USER }}
          key: ${{ secrets.SSH_PRIVATE_KEY }}
          script: |
            cd ${{ secrets.VPS_TARGET_DIR }}
            docker compose pull <service>
            docker compose up -d <service>
```

## Key Patterns

### Matrix Builds
```yaml
jobs:
  build:
    strategy:
      matrix:
        service: [frontend, api, worker]
      fail-fast: false  # Let others complete even if one fails
```

### Conditional Jobs
```yaml
jobs:
  deploy:
    if: github.event.workflow_run.conclusion == 'success'
```

### Caching
```yaml
- uses: actions/cache@v4
  with:
    path: ~/.bun/install/cache
    key: ${{ runner.os }}-bun-${{ hashFiles('bun.lock') }}
    restore-keys: |
      ${{ runner.os }}-bun-
```

### Secrets
```yaml
# All secrets in GitHub Secrets, never in code
secrets:
  SSH_PRIVATE_KEY: ${{ secrets.SSH_PRIVATE_KEY }}
  VPS_HOST: ${{ secrets.VPS_HOST }}
  ENV_FILE_PRODUCTION: ${{ secrets.ENV_FILE_PRODUCTION }}
```

## Environment Strategy

| Environment | Purpose | Deploy Method |
|-------------|---------|---------------|
| `development` | Local dev | Manual `docker compose up` |
| `staging` | Pre-production | Auto-deploy from PR branches |
| `production` | Live | Auto-deploy from main |

## Quality Gates (run order)

1. **Lint** (<1 min) — Biome/Ruff/clippy + format check
2. **Type check** (<2 min) — tsc/pyright/cargo check
3. **Unit tests** (<3 min) — fast, no external deps
4. **Build** (<5 min) — compile/transpile, build Docker images
5. **Integration tests** (<10 min) — with DB, external services
6. **Security scan** (<5 min) — CodeQL, dependency audit, secret scan
7. **Deploy** (<2 min) — SSH, pull, restart

## Deployment Pattern

1. SSH to deployment target
2. Pull latest images from container registry
3. Restart specific container (not all)
4. Health check after restart
5. Rollback if health check fails

## Anti-patterns

- ❌ **Large CI configs** — extract repeated blocks into actions. Use YAML anchors.
- ❌ **Building in deployment** — build once in CI, deploy artifact. Avoid `docker compose build` on production.
- ❌ **Hardcoded values** — use env vars, secrets, and GitHub variables.
- ❌ **Skipping lint** — lint should run first and gate everything.
- ❌ **No caching** — each run fetches deps fresh = 2x+ slower.
- ❌ **Manual deployment steps** — automate everything. If it's manual, it will be wrong.
- ❌ **Deploying untested artifacts** — run tests before build, not after.