cicd-pipeline · diff

git:20260708.e42b71f to git:20260911.3a0a7c8

3 added, 3 removed. Audit A to A.

---

name: "cicd-pipeline"

description: "Generate CI/CD workflows (GitHub Actions, or self-hosted Woodpecker via provider)"

---

# CI/CD Pipeline Generation

Generate CI/CD workflows from your Makefile targets. Defaults to GitHub Actions;

set `pipeline.provider: woodpecker` (or `both`) in `.claude/cicd.yml` to emit a

self-hosted Woodpecker `.woodpecker.yml`. For a hardened self-hosted pipeline

(secret-scan + image-security + runtime-smoke) and to scaffold the Woodpecker

server/agent, use `/cicd-woodpecker` instead.

## Steps

1. **Check for task manifest** - if `.claude/cicd_tasks.yml` exists, use it to inform pipeline generation:

```bash

if [ -f ".claude/cicd_tasks.yml" ]; then

echo "Found cicd_tasks.yml manifest - pipeline will use manifest-defined steps"

# The manifest defines plan steps (lint, test, deploy, etc.) with exact commands.

# Pipeline generation should use these commands instead of defaults.

```

When a manifest is present, the generated pipeline YAML should:

- Use the exact commands from the manifest's step definitions

- Include timeout values from the manifest

- Respect skip_if conditions as conditional steps

- Fall back to `make <target>` for steps not in the manifest

2. **Detect framework** using `lib/cicd`:

```bash

- PYTHONPATH="$PWD/lib:$HOME/Projects/claude-power-pack/lib:$PYTHONPATH" python3 -m lib.cicd detect --quiet

+ PYTHONPATH="$PWD:$HOME/Projects/claude-power-pack:$PYTHONPATH" uv run --project "$HOME/Projects/claude-power-pack" python -m lib.cicd detect --quiet

```

3. **Generate pipeline** (dry run first):

```bash

- PYTHONPATH="$PWD/lib:$HOME/Projects/claude-power-pack/lib:$PYTHONPATH" python3 -m lib.cicd pipeline

+ PYTHONPATH="$PWD:$HOME/Projects/claude-power-pack:$PYTHONPATH" uv run --project "$HOME/Projects/claude-power-pack" python -m lib.cicd pipeline

```

4. **Review output** with the user. Show the generated workflow YAML.

5. **Check for existing files** before writing:

- If `.github/workflows/ci.yml` exists, ask before overwriting

6. **Write files** if approved:

```bash

- PYTHONPATH="$PWD/lib:$HOME/Projects/claude-power-pack/lib:$PYTHONPATH" python3 -m lib.cicd pipeline --write

+ PYTHONPATH="$PWD:$HOME/Projects/claude-power-pack:$PYTHONPATH" uv run --project "$HOME/Projects/claude-power-pack" python -m lib.cicd pipeline --write

```

7. **Report results**:

```

## CI Pipeline Generated

Framework: {framework} ({package_manager})

Files created:

.github/workflows/ci.yml - CI pipeline with lint, test, build

Triggers: push to main, pull requests

Targets: make lint, make test, make typecheck (if available)

To view: cat .github/workflows/ci.yml

```

## Notes

- Workflows use `make <target>` as steps (not direct tool commands)

- This keeps CI in sync with local development commands

- Caching is included for package managers (uv, npm, cargo, go)

- Matrix builds are configured from `.claude/cicd.yml` if present

- Configure pipeline settings in `.claude/cicd.yml`:

```yaml

pipeline:

provider: github-actions # github-actions | woodpecker | both

branches:

main: [lint, test, typecheck, build, deploy]

pr: [lint, test, typecheck]

matrix:

python: ["3.11", "3.12"]

```

- Self-hosted Woodpecker: set `provider: woodpecker` (or `both`) and see

`/cicd-woodpecker` + `docs/skills/woodpecker-ci.md` for the hardening stages.