CI/CD Pipeline Review
Deep review of CI/CD pipelines — GitHub Actions, secrets management, caching strategy, build optimization, deployment safety, and security hardening.
- License 许可
- MIT
- 最近更新
- 2026-10-09
- Stars
- ★ 1
- Forks
- 0
01 前言 Front Matter
- name
- CI/CD Pipeline Review
- license
- MIT
- version
- 2026.3.1
- triggers
- { "lenses": [ "security", "logic" ], "keywords": [ "ci", "cd", "github actions", "pipeline", "workflow", "deploy" ] }
- description
- Deep review of CI/CD pipelines — GitHub Actions, secrets management, caching strategy, build optimization, deployment safety, and security hardening.
03 SKILL.md 正文
When reviewing CI/CD configuration, apply the following checks.
Secrets Management
Flag secrets hardcoded in workflow files — use repository or environment secrets. Flag echo ${{ secrets.TOKEN }} — this prints the secret to logs even though GitHub masks known values; log masking is not guaranteed for derived values.
Verify secrets are scoped to the minimum required environment (production secrets only available in the production environment). Flag workflows that pass secrets to third-party actions via environment variables without reviewing the action source.
Check that GITHUB_TOKEN permissions are restricted with permissions: block — the default token has write access to the repository. Flag permissions: write-all.
Action Security
Flag third-party actions pinned to a branch (uses: owner/action@main) instead of a commit SHA (uses: owner/action@abc123). Branch references can be silently updated to malicious code. Verify high-privilege actions (deploy, publish, release) use SHA-pinned versions.
Flag pull_request_target trigger with actions/checkout of the PR branch — this executes untrusted PR code with write permissions. Flag workflow_dispatch without input validation.
Caching
Verify dependency caching is enabled for package managers (npm, pnpm, pip, cargo). Flag workflows that install dependencies from scratch on every run. Check that cache keys include the lockfile hash to invalidate on dependency changes.
Flag Docker layer caching that uses type=gha without scope — this can leak layers between unrelated workflows. Verify build caches are scoped to the branch.
Build Optimization
Flag sequential steps that could run in parallel using a matrix strategy. Flag duplicate npm install or pip install across multiple jobs — use a shared setup job with artifact upload. Check that --frozen-lockfile (pnpm) or --ci (npm) is used — installing without lockfile verification can produce non-reproducible builds.
Flag test suites that run the entire suite on every PR — use test splitting or affected-file detection for large repositories.
Deployment Safety
Verify production deployments have a manual approval step (environment protection rules). Flag direct pushes to production without a staging/preview deployment. Check that rollback procedures exist — deployment workflows should support reverting to the previous version.
Flag deployments that do not verify health after deploying (no smoke test or health check step). Verify database migrations run before the new application version starts, not after.
Workflow Structure
Flag continue-on-error: true on critical steps (tests, security scans, linting) — failures should block the pipeline. Flag missing timeout-minutes on jobs — default timeout is 6 hours, which wastes resources on hung jobs.
Verify status checks are required for the default branch — flag repositories where PRs can be merged without passing CI. Check that branch protection rules enforce up-to-date branches before merging.
For each finding, report: the workflow file and step, the specific CI/CD pattern violated, and the recommended fix.
04 同仓库更多 Skills
Vue.js Review
Deep review of Vue.js applications — Composition API patterns, reactivity traps, component design, Pinia state management, and performance pitfalls.
Unity Review
Deep review of Unity C# code — object pooling, Update allocation, physics patterns, UI performance, serialization, and mobile optimization.
TypeScript Review
Deep review of TypeScript code — type safety, generics patterns, strict mode compliance, type narrowing, module boundaries, and common escape hatches.
Test Conventions
Testing patterns and conventions for code review. Covers naming, AAA structure, boundary testing, mock vs real dependencies, and coverage expectations.
Terraform Review
Deep review of Terraform/OpenTofu configuration — state management, module design, security hardening, drift prevention, and production safety patterns.
Swift Review
Deep review of Swift code — memory management, concurrency patterns, optionals, protocol design, SwiftUI lifecycle, and iOS platform patterns.