Quality Gates

Definition of Done

• All tests pass (cargo test)
• No clippy warnings (cargo clippy -- -D warnings)
• Code formatted (cargo fmt --check)
• No hardcoded paths or credentials
• Public APIs have doc comments
• Error cases handled, not ignored

Build & Test Commands

cargo build                # Build
cargo test                 # Run all tests
cargo clippy               # Lint
cargo fmt --check          # Format check
cargo check                # Type check only

• ALWAYS run cargo test after code changes
• ALWAYS run cargo check before committing
• ALWAYS run cargo clippy before opening PRs

Hook Testing

• After modifying any hook script in .claude/hooks/, run bash tests/hooks/run_hook_tests.sh
• When creating test files, be aware that protect-hooks.sh may block writes to hook-adjacent paths — if blocked, report to team lead immediately
• Ensure git test environments have user.email and user.name configured

Agent Output Schemas

When spawning agents, expect structured output:

• coder: { files_changed: [], tests_added: [], tests_passing: bool }
• tech-lead: { status: APPROVED|BLOCKED, critical: [], warnings: [] }
• reviewer: { issues: [], suggestions: [], approved: bool }
• qa: { tests_run: int, tests_passed: int, edge_cases: [] }

Review Pipeline

1. Coder completes implementation
2. cargo test + cargo clippy runs (hook, deterministic)
3. Tech-lead reviews code (spawned as subagent)
4. If BLOCKED → coder fixes and re-submits
5. If APPROVED → orchestrator commits
6. User reviews via git-review (review gate hook)

CI/CD Integration

• GitHub Actions should run: cargo test, cargo clippy, cargo fmt --check
• The review gate (git-review gate check) can run in CI as a required check
• Hook scripts in .claude/hooks/ are for local Claude enforcement only — not CI
• CI checks are deterministic and repeatable; hook checks add agent-specific rules