More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    GitHub Actions Workflow Templates

    1

    Design GitHub Actions pipelines from first principles: scoped triggers, hardened jobs, measured caching, sane matrices, gated deploys, and calm failure triage.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    GitHub Actions Workflow Templates

    GitHub Actions Workflow Templates

    Example session with this skill installed

    Set up CI for my Node app. It's a single repo, tests take about six minutes, dependencies take forever to install, and right now there's nothing — I just merge and hope. I don't need deploys yet, just checks I can trust on every pull request.

    • Read your context and instructions
    • Compiled the github actions workflow

    Design, from the four pipeline obligations down.

    One workflow, ci.yml, triggered on pull_request, plus a push run on main so trunk breakage is visible instead of discovered by the next PR. Top-level permissions: {}; the single job grants itself contents: read and nothing else — a compromised dependency in the test run cannot reach packages, releases, or deployments.

    Job shape: checkout pinned at v4, setup-node at v4 with cache: npm, then npm ci (lockfile-deterministic, never npm install), lint, and npm run test:ci. timeout-minutes: 10 — the suite normally finishes in six, so a hang dies in ten minutes instead of billing for six hours. On caching: your installs are the expensive kind, so the cache clearly pays; it is keyed on the lockfile hash by the setup action, and the expected saving is written into the rationale note to verify against the next three runs.

    Deliberately left out, with reasons. No matrix: the app ships on one Node version in production, and testing versions it never runs on is decoration — if that changes, the matrix template adds legs in one block. No deploy job: you said not yet, and when it arrives it belongs in a separate tag-triggered workflow behind a production environment with a required reviewer, so a merge can never reach prod by itself. No chat notification on PR checks: a channel that fires on every PR trains everyone to ignore it; notifications belong on trunk breakage and deploys.

    Settings checklist delivered with the YAML: mark the test job as a required status check on main in branch protection; no repository secrets are needed at all — the scoped built-in token covers CI. Readiness rubric on the delivered design: 16/18, soft rows being OIDC (nothing to federate yet; no cloud calls in CI) and notifications (scoped away from PRs as a recorded decision).

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Most CI pipelines are accumulated, not designed — copied from a starter, patched under deadline, and quietly expensive ever since. This skill designs them from first principles, starting from the four things a pipeline owes you: fast signal on pull requests, reproducible builds, a controlled blast radius, and logs that point at the cause when something breaks.

    The method works through the decisions in the order they matter. Trigger strategy first, from a decision table: pull requests gate, version tags release, cron sweeps, and manual dispatch handles deploys and rollbacks — nothing runs on events that don't need it, and path filters keep docs changes from burning monorepo compute. Job anatomy next: permissions emptied at the top and granted narrowly per job, timeouts on everything, actions pinned to versions, lockfile installs — each choice shipped with its reason, because a template you don't understand becomes superstition.

    Caching is treated as an equation, not a ritual: measure the install step first, skip the cache entirely under a minute, and key what you do cache on the lockfile hash, with the measured saving recorded. Matrices are sized by what actually differs, with include/exclude discipline and an eye on the multiplier. Container releases build one immutably tagged image (semver plus commit SHA from git state) that is promoted through environments rather than rebuilt per environment. Deployments sit behind GitHub environments with required reviewers, environment-scoped secrets, and rollout verification, so "what's in prod?" always has an answer.

    It also covers reusable workflows across repos (pinned at the caller, named secrets only — never secrets: inherit to outside code), security scanning wired into the security tab with thresholds people won't learn to ignore, OIDC over long-lived cloud keys, and a four-step failure triage order for when a run goes red. Every delivery is scored on a nine-row pipeline readiness rubric and includes the settings checklist — environments, branch protection, secret names — that lives outside the YAML.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for DevOps & Cloud.

    The templates and decision rules are GitHub Actions-specific; the principles transfer to other CI systems but the YAML, permission model, and caching mechanics do not. Pipeline design cannot fix slow or flaky tests themselves — the triage protocol will identify them, and caching can hide some of the cost, but a 40-minute test suite is a test problem, not a workflow problem. Organization-level features (required workflows, runner groups) vary by GitHub plan, and the skill flags where a design assumes them.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Money-back guarantee

    Doesn't do what its page promises? You get your money back.

    Trust & safety

    Security scan

    Security scan passed 24 days ago

    Software
    Python

    Permissions

    No special permissions declared or detected

    Listed24 days ago
    Updated21 days ago

    What's inside

    Frequently Asked Questions

    No. The YAML in the skill is the worked reference, but the deliverable is a design made from your stack, repo layout, and release model — with a rationale note for every non-obvious choice and a settings checklist for the parts that live outside the YAML. A pasted starter is exactly the failure mode this skill exists to replace.

    Yes — that's the second entry point. The readiness rubric scores the existing pipeline across nine rows (pinning, permissions, timeouts, trigger scope, caching, determinism, deploy gate, secrets, signal), the caching equation finds where the time goes, and the triage order separates real failures from flakes before you change anything.

    Both, with a boundary: deployments are designed as separate tag-triggered workflows gated by GitHub environments with required reviewers and environment-scoped secrets. If nobody can name who approves production, the skill refuses to write the deploy job — an ungated deploy is an incident scheduled in advance.

    New in DevOps & Cloud