INTERMEDIATE · 8 MODULES · 64 LESSONS
Claude Code in CI/CD Pipelines
Put Claude Code to work on every pull request in your pipeline, reviewing, fixing, and preparing releases with nobody at the keyboard, under tight limits on permissions and spend.
- For you if
- You already use Claude Code in your terminal and want it working on every PR without calling it by hand.
- By the end
- You'll know how to set up automated reviews in GitHub and GitLab, write your own pipeline agent, and protect it from malicious PRs.
- You need
- Claude Code on a paid plan or API key, GitHub Actions or GitLab CI experience, and a test repo.

Curriculum
01Claude Code in GitHub Actions
- Introduction: the night-shift contractor
- Installing with `/install-github-app` and by hand
- Your first `@claude` workflow, line by line
- Authentication and secrets
- Automation mode: `prompt`, triggers and `claude_args`
- Outputs: `structured_output`, job summary and artifacts
- Costs: minutes, tokens and brakes
- Project: Claude on call at Colibrí
03claude -p in GitLab CI/CD and other CI systems
- Module 3 introduction: claude -p outside GitHub
- claude -p in CI: bare mode, JSON output and exit codes
- GitLab CI/CD for a Claude job: stages, jobs, rules, variables and artifacts
- The GitLab pipeline with Claude: the official job, merge requests and credentials
- Permissions for unattended runs
- claude -p from Python and TypeScript scripts
- Other CI systems: the common recipe
- Project: porting Colibrí's review bot to GitLab
04The Claude Agent SDK in your own pipelines
- Module 4 introduction: the Claude Agent SDK in your own pipelines
- Choosing: action, claude -p, Agent SDK, Messages API or Managed Agents
- The Agent SDK in Python inside a job
- The Agent SDK in TypeScript inside a job
- Permissions, hooks and budgets in the SDK
- Structured results and exit codes
- The no-agent alternative: one Messages API call
- Project: broken-CI triage agent for Colibrí
05Securing Claude in CI
- Module 5 introduction: securing Claude in CI
- Threat model of an agent in CI
- Least privilege: job permissions, identities and tokens
- Secrets and short-lived credentials
- Prompt injection from PRs, forks, comments and logs
- Budget guards for unattended runs
- Auditing and observability
- Project: hardening Colibrí's workflows
06Delivery automation
07Security scanning and rollback
- Module 7 introduction: security scanning and rollback
- Security scanning with Claude: the dedicated action, /security-review and your own prompt
- Integrating deterministic tools and triaging with Claude
- Automatic rollback with deterministic triggers
- Diagnosis after the rollback
- Findings as SARIF and annotations
- Incident summary and postmortem draft
- Project: security gate and rollback for Colibrí
08Project: Colibrí's complete pipeline
- Project introduction: Colibrí's complete pipeline
- End-to-end pipeline design
- Stages 1 to 3: PR review, scanning and description
- Stages 4 to 6: release, readiness and deploy with approval
- Failure paths and observability
- Hardening and security review
- Documentation, cost and retrospective
- Final delivery and rubric
Reviews
What students say
These reviews are from enrolled students who completed at least 50% of the course. We moderate reviews only on content grounds (spam, offensive language, personal data), never for being critical or negative.
No approved reviews yet.
Be the first to share your experience!