← Dashboard

📋 CI/CD Skills Engine

GitHub CI/CD workflows as Claude Skills — sandbox execution, idempotency, dry-run safety, and an LLM auto-router that picks skills from a natural-language query.

Either type a free-form query, or tick the skill chips below — at least one form of intent is enough. With a query, the router LLM plans + chains skills. Without one, the router runs the included chips (or the default dependency-audit + security-scan pair) and skips the planning LLM call entirely (pure cost win).
vulns + leaked creds leaked secrets dep audit ship-readiness full health check tests + lint ⊘ clear query
👍 Include hints (run these)
lint-and-test dependency-audit security-scan build-and-release
When the query is empty, the included chips become the plan directly — no LLM plan call. Including build-and-release counts as explicit release intent.
🚫 Exclude hints (skip these)
lint-and-test dependency-audit security-scan build-and-release
Hard block — the router will refuse to schedule these even mid-loop.
Choose a preset or type any model id, e.g. qwen/qwen3-next-80b-a3b-instruct.
Set this when a free-text model prefix is ambiguous.
openrouter.ai/credits
Server-bundled key may be exhausted — get your own free key at build.nvidia.com
⚠️ LIVE MODE — will create tags/releases
Only applies when build-and-release runs. All read-only skills always run live.
— —