Ci cd pipeline
Self-hosted AI coding factory — sandboxed agents deliver tickets to merged code, gated by a human in a dashboard. Local-first, cost-transparent, human-in-the-loop.
npx -y skills add tmj-90/gaffer --skill ci-cd-pipelineAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Use when setting up CI for a new project, refactoring existing pipelines, migrating between platforms, or standardising deployment workflows. Triggers on "set up CI/CD", "GitHub Actions", "GitLab CI", "pipeline for this repo", "deployment workflow", or "why is CI slow".
SKILL.md
3.4 KB, as published. Nobody here has run it
Build pragmatic CI/CD pipelines
A pipeline that runs in 2 minutes and deploys reliably beats one that does everything and breaks weekly. Detect the real stack; emit stages that match the project's actual commands.
Canonical pipeline stages
lint → test → build → scan → deploy-staging → smoke-test → deploy-prod
Not every project needs all stages. Drop what adds no signal; never skip test and lint.
Stack detection signals
| File present | Inferred stack |
|---|---|
package.json + tsconfig.json | TypeScript/Node |
pom.xml | Java/Maven |
build.gradle | Java/Gradle |
requirements.txt / pyproject.toml | Python |
go.mod | Go |
Cargo.toml | Rust |
Dockerfile | Container build needed |
.terraform/ or *.tf | Terraform plan/apply stage |
Steps
- Detect the stack. Read
package.json,pom.xml, lockfiles, andDockerfile— do not guess. Extract the actual test, lint, and build commands from the project's own scripts. - Choose the minimal stage set. Map detected commands to pipeline stages. If a stage has no command, omit it — a placeholder stage that always passes adds noise.
- Configure caching. Cache dependency directories keyed by lockfile hash (e.g.
pnpm-lock.yaml,go.sum). A cache miss should still produce a correct build. - Add secrets hygiene. All credentials via CI secret store — never hardcoded. Mask secrets in logs. Principle of least privilege for deploy tokens.
- Emit the pipeline file. GitHub Actions (
.github/workflows/ci.yml) or GitLab CI (.gitlab-ci.yml) depending on the platform. Validate YAML syntax before committing. - Verify. Trigger the pipeline on a feature branch; confirm all stages pass; record timings; submit for review.
Build / Test
- Validate YAML with the platform's own linter (
actionlintfor GitHub Actions,gitlab-ci-lintfor GitLab) before push. - Confirm cache restores by running the pipeline twice — second run should be significantly faster.
- Deploy stages: dry-run (
--dry-runor preview) first; then gate production deploy on staging smoke-test pass.
Review checklist
- Stages match actual project commands — no placeholder steps, no undocumented scripts.
- Caching wired — lockfile-keyed; cache miss still produces correct build.
- No secrets in YAML — all credentials from CI secret store; masked in output.
- Minimal stage set — every stage has a real command and a reason to exist.
- Branch protection respected — PR pipeline required-to-pass before merge on default branch.
Rules
- Always read the project's own
package.json/Makefile/pom.xmlfor the real commands — never invent them. - Secrets from the CI secret store only; rotation policy documented.
- Fail fast: put linting before testing; put cheap checks before expensive ones.
Capture lore
Pipeline structure, deploy targets, secret names, and environment promotion policy are high-value lore. Call suggest_lore with tags: [ci-cd, pipeline, deploy] when you learn them.
Gives 0 of the 12 instructions most ci cd skills give
Counted across 392 of the 394 authors here whose files we hold, read 2026-08-06
- pin third-party actions to full commit SHAsin 33 of 392
- cache dependencies appropriatelyin 24 of 392, across 12 files
- optimize pipelines exceeding ten minutesin 20 of 392, across 6 files
- enforce all quality gates before mergein 20 of 392, across 7 files
- Configure branch protection rulesin 19 of 392, across 5 files
- use environments for deployment trackingin 19 of 392, across 7 files
- implement manual gates for productionin 19 of 392, across 7 files
- implement security scanningin 18 of 392, across 5 files
- fix failing code instead of disabling checksin 18 of 392, across 4 files
- use CI/CD variables for secretsin 18 of 392, across 6 files
- move checks upstream in the pipelinein 17 of 392, across 3 files
- use specific image tagsin 17 of 392, across 5 files
Said here and by no other author read
- extract commands from project files
- choose the minimal stage set
- drop stages that add no signal
- never skip the test and lint stages
- load credentials from the CI secret store
- put cheap checks before expensive ones
Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once.