cicd

CI/CD Setup

Generate a self-contained .github/workflows/ci.yml tailored to the project's language stack. The workflow runs on every PR and on pushes to main, with lint + test + (optional) build gates.

Self-contained. No external reusable workflows, no shared repos. Each project owns its CI file outright.

Idempotent. If .github/workflows/ci.yml already exists, do not overwrite. Tell the user it exists, show its language coverage, and ask whether to reconcile (add missing language jobs) or leave it.

Steps

1. Detect the language stack

Read the repo root and identify which of the supported languages are in play. Look for these markers (a project can be polyglot, generate jobs for every language found):

Language	Marker files
Node / TypeScript	package.json (check engines.node, scripts, lockfile for npm/pnpm/yarn/bun)
Python	pyproject.toml, requirements.txt, setup.py, Pipfile
Go	go.mod
Rust	Cargo.toml

If none of these are present, tell the user the project's language stack isn't recognized and stop. Don't write a workflow that won't work.

If multiple languages are detected, generate a workflow with one job per language. The jobs run in parallel.

2. Check for existing workflow

If .github/workflows/ci.yml exists:

• Read it.
• Identify which languages it covers (look for setup-node, setup-python, setup-go, dtolnay/rust-toolchain).
• If it already covers everything detected in step 1, tell the user "CI workflow already in place for [languages]; nothing to do" and stop.
• If it's missing language(s), offer to add jobs for the missing ones. Don't rewrite existing jobs.

If no workflow exists, proceed to generate one.

3. Generate the workflow

Assemble .github/workflows/ci.yml from the snippets below. Use a single name: CI header and one job per detected language. Each job runs on ubuntu-latest.

Common header:

name: CI

on:
  pull_request:
  push:
    branches: [main]

jobs:

Node / TypeScript job. Detect the package manager from the lockfile (pnpm-lock.yaml → pnpm; yarn.lock → yarn; bun.lockb → bun; otherwise → npm). Adjust the install + run lines accordingly. Use --if-present for scripts so the job doesn't fail when a script isn't defined.

  node:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
          cache: 'npm'  # change to pnpm/yarn if detected
      - run: npm ci
      - run: npm run lint --if-present
      - run: npm run typecheck --if-present
      - run: npm test --if-present
      - run: npm run build --if-present

Python job. Default to pip + pyproject.toml. If poetry.lock is present use poetry; if uv.lock is present use uv. Run ruff if ruff appears in pyproject deps or ruff.toml/.ruff.toml exists. Run pytest unconditionally, the project will fail-fast if no tests exist, which is fine.

  python:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.12'
          cache: 'pip'
      - run: pip install -e ".[dev]" || pip install -r requirements.txt
      - run: ruff check .
        if: ${{ hashFiles('ruff.toml', '.ruff.toml') != '' || contains(fromJSON('[true]'), false) }}
      - run: pytest

(If the ruff conditional is too clever, just emit a plain ruff check . || true, getting a real ruff signal matters more than perfect conditional gating.)

Go job. Read go.mod for the module's Go version directive and use it.

  go:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with:
          go-version-file: 'go.mod'
      - run: go vet ./...
      - run: go test ./...
      - run: go build ./...

Rust job.

  rust:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: dtolnay/rust-toolchain@stable
        with:
          components: clippy, rustfmt
      - uses: Swatinem/rust-cache@v2
      - run: cargo fmt --check
      - run: cargo clippy -- -D warnings
      - run: cargo test
      - run: cargo build

4. Write the file

Create .github/workflows/ if it doesn't exist, then write the assembled YAML to .github/workflows/ci.yml. Don't add a trailing newline beyond the standard single one.

5. Tell the user what you did

Report:

• Languages detected and the corresponding jobs added.
• Package manager detected for Node (npm/pnpm/yarn/bun), important so they know what npm ci resolved to.
• File path written.
• Whether anything was skipped (e.g., workflow already existed and covered everything).
• Next step: commit, push, and open a PR to see the workflow run.

If the project uses an unusual setup (custom test commands, non-standard layout, monorepo with multiple packages), flag it, the generated workflow assumes the project root holds the package config. Suggest manual edits if that doesn't hold.

What this skill does NOT do

• Coverage thresholds (each language ecosystem has its own tooling; opt in per-project).
• Security scans (Trivy, gitleaks). Add them later if the project's threat model needs it.
• Deploy steps. Deploy belongs in a separate workflow keyed to the project's hosting platform, not in ci.yml.
• Mutation testing. Niche.
• AI-powered failure feedback. Cool, but cross-cutting concern.

Keep the generated workflow lean. Projects grow into more later; they shouldn't start there.