Testing

Running Tests

npm test              # all 66 tests, 9 suites
npm run test:watch    # watch mode with re-run on change

Test Suites

Suite	Tests	What It Covers
config.test.ts	10	Config loading, validation, env var resolution, defaults
airs-client.test.ts	5	SDK wrapper, prompt/response scanning, block verdicts
scanner.test.ts	10	Observe/enforce/bypass modes, UX messages, fail-open
code-extractor.test.ts	10	Fenced, indented, heuristic extraction, mixed content
circuit-breaker.test.ts	7	State transitions, cooldown, probe success/failure
dlp-masking.test.ts	8	Enforcement priority, content masking
logger.test.ts	5	JSON Lines output, content stripping, directory creation
log-rotation.test.ts	3	Size threshold, rotation, missing file handling
hooks-integration.test.ts	8	End-to-end Cursor JSON contract (tsx + compiled JS)

Integration Tests

The hooks-integration.test.ts suite runs actual hook scripts with piped JSON:

// Runs: echo '{"prompt":"..."}' | npx tsx src/hooks/before-submit-prompt.ts
// And:  echo '{"prompt":"..."}' | node dist/hooks/before-submit-prompt.js

These verify the real Cursor contract end-to-end, including JSON parsing, config loading, fail-open behavior, and correct output format.

Mocking

• AIRS API: Not mocked in integration tests -- hooks fail-open when the API is unreachable, which validates the fail-open design
• SDK: Mocked via vi.mock("@cdot65/prisma-airs-sdk") in unit tests (airs-client.test.ts, scanner.test.ts)
• File system: Tests use temporary directories (test/.tmp-*) cleaned up in afterEach

Adding Tests

1. Create test/<module>.test.ts matching the source file
2. Use vitest's describe/it/expect API
3. Clean up any temp files in afterEach
4. For integration tests, use execSync to run hooks with piped JSON