A Claude skill that sets up automated CI/CD for any project in one command.
Auto-detects your language and framework architecture (Next.js, React, Vue, Angular, NestJS, Python, Go, Rust, Docker), writes a tested GitHub Actions workflow, and wires up secrets, environments, and dependency updates — the way a careful senior engineer would, not a copy-pasted tutorial.
Point it at a project and it will:
- Intelligently detect language & framework — Next.js, React, Vue, Angular, NestJS, Express, Python, Go, Rust, and Docker containers.
- Differentiate Frontend vs. Backend — Recommends Vercel / Cloudflare for frontend web apps, and Railway / Render / SSH for backend APIs and services.
- Generate a real workflow — a
testjob gates adeployjob; nothing reaches production without passing tests first. - Set up Dependabot — weekly dependency + GitHub Actions version updates, for free.
- Tell you exactly what to do next — the precise secret names to add, and which GitHub Environment to create.
- Audit an existing workflow —
--validatechecks any.github/workflows/*.ymlfor missingpermissions:, ungated deploy jobs, and unpinned@main/@masteractions;--strictexits non-zero for CI use.
Run this command in your project terminal to install the CI/CD skill into your IDE:
bash <(curl -fsSL https://raw.githubusercontent.com/khemratechconsulting/github-actions-setup/main/skills/github-actions-setup/scripts/setup.sh)We take script execution security seriously:
- Transparent Source Code: The script is 100% open source shell script inspectable directly at
skills/github-actions-setup/scripts/setup.sh. - Zero-Dependency & No Elevating Privileges: Does not require
sudo, root permissions, or binary downloads. It only manipulates.github/workflows/and local.agents/skills/text files inside your repository workspace.
It auto-detects your IDE's skill directory:
.agents/skills/(Cursor, GitHub Copilot, Gemini CLI, Amp, Cline, Warp).agent/skills/(Antigravity).claude/skills/(Claude Code).void/skills/(Void IDE).windsurf/skills/(Windsurf).kiro/skills/(Kiro).codex/skills/(Codex / OpenAI).qoder/skills/(Qoder)
Once installed, simply talk to your AI assistant inside your IDE in natural language:
- 💬 "Set up GitHub Actions CI/CD for this project."
- 💬 "Configure automated deployment to Vercel (or Cloudflare / Railway / AWS)."
- 💬 "Audit our current
.github/workflows/ci-cd.ymlfor security issues." - 💬 "Add Dependabot for automated dependency security updates."
Your AI assistant will read the local skill rulebook and execute the setup.sh automation script directly for you!
Tip: Add the skill directory to your project's .gitignore if you prefer not to commit internal skill documentation into your repo history.
See skills/github-actions-setup/SKILL.md for full trigger details.
| Template | Covers | Auth |
|---|---|---|
ci-only.yml |
Tests only, matrix across two Node versions | — |
ci-cd-aws.yml |
Test → deploy to S3/CloudFront | GitHub OIDC, no static keys |
ci-cd-vercel.yml |
Test → PR previews → production deploy | Static token |
ci-cd-cloudflare.yml |
Test → deploy to Cloudflare Pages (Wrangler) | Static, scoped API token |
ci-cd-railway.yml |
Test → deploy via Railway CLI | Static project token |
ci-cd-render.yml |
Test → trigger a Render deploy hook | Static hook URL |
ci-cd-generic-server.yml |
Test → SSH deploy + process restart | Static SSH key |
AWS is the only target here with GitHub OIDC federation — no long-lived credentials stored anywhere. Cloudflare, Railway, and Render don't support OIDC from GitHub yet, so those three (and Vercel/server) rely on a static secret; scope it as narrowly as the platform allows and rotate it periodically.
Full write-up of the security and efficiency choices baked into these: skills/github-actions-setup/references/best-practices.md.
The free tier above covers the three most common deploy targets and is a complete toolkit on its own.
| Tier | Price | Adds |
|---|---|---|
| Free | $0 | Everything listed above |
| Supporter | $1.99/mo | Priority email support, early access to new templates |
Get either from the pricing section on the site.
Issues and PRs welcome — see CONTRIBUTING.md. Found a security issue? See SECURITY.md instead of opening a public issue.
MIT — see LICENSE.