AG-CI
The CI/CD & Quality Agent specializes in continuous integration pipelines, test infrastructure, and code quality automation. AG-CI keeps the CI green and fast, ensuring every commit meets quality standards.
Capabilities
- CI/CD Pipeline Setup: GitHub Actions, GitLab CI, and other CI platforms
- Test Framework Configuration: Jest, Vitest, Pytest, Playwright, Cypress setup
- Linting & Formatting: ESLint, Prettier, Black configuration
- Type Checking: TypeScript, mypy, and other type systems
- Code Coverage: Coverage threshold enforcement and reporting
- Security Scanning: SAST, dependency checks, vulnerability scanning
- E2E Testing: End-to-end test automation and flaky test detection
- Build Optimization: Caching, parallelization, job splitting
- CLAUDE.md Maintenance: Documentation of CI/test practices and conventions
When to Use
Use AG-CI when you need to:
- Set up or optimize CI/CD pipelines
- Configure test frameworks and harnesses
- Implement linting, formatting, or type checking
- Set up code coverage tracking and enforcement
- Optimize build times
- Fix flaky tests
- Implement security scanning
- Document testing practices in CLAUDE.md
How It Works
- Context Loading: Agent reads expertise file and checks CI configuration
- Analysis: Reviews status.json for test-related stories and CI issues
- Execution: Sets up workflows, test frameworks, quality tools
- Verification: Runs CI pipeline on feature branch to ensure it passes
- Documentation: Updates CLAUDE.md with new testing patterns and conventions
- Coordination: Updates status.json and communicates via agent bus
Example
# Via /babysit
/agileflow:babysit
> "We need to set up a CI pipeline with unit and E2E tests"
# Or directly invoke
/agileflow:setup-tests
# AG-CI will:
# 1. Detect project type (Node.js, Python, etc.)
# 2. Recommend test frameworks
# 3. Set up test infrastructure
# 4. Configure CI workflows
# 5. Document patterns in CLAUDE.mdKey Behaviors
- CI must stay green - Cannot merge failing tests
- Keep builds fast - Target under 5 minutes for unit/lint, under 15 minutes full suite
- Fail fast approach - Catch regressions immediately, not at PR review
- No disabled tests without approval - Coverage thresholds are enforced
- Security scanning enabled - Every CI pipeline includes security checks
- CLAUDE.md updated proactively - Test patterns documented for future development
- Context Preservation: Uses compact_context (priority: high) to maintain CI pipeline focus during long conversations, preserving build strategy and coverage targets through context compaction
Compact Context Configuration
The CI agent uses high priority compact_context to ensure build integrity stays in focus:
compact_context:
priority: high
preserve_rules:
- "LOAD EXPERTISE FIRST: Always read packages/cli/src/core/experts/ci/expertise.yaml"
- "CI MUST STAY GREEN: Cannot merge failing tests (non-negotiable)"
- "KEEP BUILDS FAST: Target under 5 minutes unit/lint, under 15 minutes full suite"
- "COVERAGE THRESHOLDS: Enforce minimums, no disabled tests without approval"
- "SECURITY SCANNING: Every CI pipeline includes security checks"
- "FAIL FAST APPROACH: Catch regressions immediately"
state_fields:
- current_story
- ci_pipeline_status
- coverage_targets
- flaky_tests_found
- build_optimization_progressThis ensures CI-critical rules (green builds, coverage thresholds, security scanning) and current state (pipeline status, coverage gaps, flaky tests) remain in focus through context compaction.
Tools Available
This agent has access to:
- Read: Access workflow files and test configurations
- Write: Create new test files and workflow definitions
- Edit: Update existing CI configurations
- Bash: Execute CI commands and test runners
- Glob: Find test files and workflow patterns
- Grep: Search for test patterns and coverage issues
Core Responsibilities
- Keep CI green and fast
- Ensure at least one test per story validates acceptance criteria
- Maintain test coverage thresholds
- Configure linting, formatting, type checking
- Set up security scanning
- Document testing practices in docs/02-practices/
- Update status.json after each status change
- Append coordination messages to agent bus
- Fix flaky tests and CI issues
- Optimize build times
Quality Standards
Before marking work complete, AG-CI ensures:
- CI runs successfully on the feature branch
- Unit/lint jobs complete in under 5 minutes
- Integration/E2E tests run in parallel where possible
- Failed tests provide clear, actionable error messages
- Coverage reports generated and thresholds met
- Required checks configured on protected branches
- Flaky tests identified and fixed (or tracked)
- Security scanning enabled
- Secrets accessed via GitHub secrets, not hardcoded
Related Agents
AG-DEVOPS- Coordinate on build optimization and performanceAG-UI- Provide component test setup and accessibility testingAG-API- Provide integration test setup and test databaseMENTOR- Report on test infrastructure gaps
Slash Commands
AG-CI can directly invoke these commands:
/agileflow:research:ask TOPIC=...- Research test frameworks and CI platforms/agileflow:ai-code-review- Review CI configuration before marking in-review/agileflow:impact-analysis- Analyze impact of CI changes on build times/agileflow:adr-new- Document CI/testing decisions/agileflow:tech-debt- Document flaky tests and slow builds/agileflow:board- Visualize story status/agileflow:status STORY=... STATUS=...- Update story status
Performance Targets
AG-CI maintains these performance standards:
- Unit/lint jobs: under 5 minutes
- Full test suite: under 15 minutes
- Build time trend: Stable or improving month-over-month
- Flaky tests: Zero in main branch
CLAUDE.md Maintenance
AG-CI proactively updates CLAUDE.md with CI/test patterns when:
- Setting up CI/CD pipeline for the first time
- Adding new test frameworks or tools
- Establishing testing conventions
- Configuring code quality tools
- Discovering project-specific testing best practices
This ensures future development uses consistent patterns.
Verification Protocol
AG-CI follows the Session Harness system to prevent breaking functionality:
- Pre-Implementation: Checks baseline test status and CI configuration
- During Work: Runs incremental tests, validates CI passes
- Post-Implementation: Verifies CI pipeline passes before marking complete
- Story Completion: Can ONLY mark "in-review" if
test_status: "passing"
See the Session Harness Protocol for complete details.