unit-test

Run tests (unit tier unless --all), emit proof files via feature-scoped overwrite, and report coverage per feature.

Usage

purlin:unit-test [feature]      Run tests for a specific feature (unit tier)
purlin:unit-test                Run all unit-tier tests
purlin:unit-test --all          Run all tests across all tiers

Step 1 — Detect Test Framework

Read .purlin/config.json for test_framework. If set to a specific framework, use it. If "auto" or missing, detect from project files using the same heuristics as purlin:init Step 3 (see references/supported_frameworks.md for the full detection logic).

Step 2 — Run Tests

# pytest (unit tier)
pytest -m "not integration"

# pytest (all tiers with --all)
pytest

# jest (unit tier)
npx jest --testPathPattern="unit"

# jest (all tiers with --all)
npx jest

The proof plugins (scripts/proof/pytest_purlin.py, scripts/proof/jest_purlin.js, scripts/proof/shell_purlin.sh) emit <feature>.proofs-<tier>.json next to the spec file. This is a feature-scoped overwrite: each run replaces all proof entries for the tested feature in that tier file, preserving entries from other features.

Proof File Freshness Check

After tests run, before reporting results, verify that proof files (*.proofs-*.json) were modified AFTER the test command started. If proof files are older than the test run or don't exist, the proof plugin didn't emit — something went wrong. Report:

WARNING: Proof files were not updated by the test run. The proof plugin may not be loaded.
→ Check: is the proof plugin registered in conftest.py / jest.config.js?
→ Run: purlin:init --force to re-scaffold the proof plugin

Never write proof JSON files directly. Only the test framework plugin writes proof files.

Step 3 — Report Coverage

Call sync_status after tests complete. Display the full result. This is not optional — without sync_status, the agent doesn't know if coverage is complete.

Test results:
  auth_login: PASSING (3/3 rules proved)
  user_profile: PARTIAL (1/2 rules proved)
    RULE-2: NO PROOF → write a test with @pytest.mark.proof("user_profile", "PROOF-2", "RULE-2")
    → PARTIAL means more tests needed to reach PASSING.
  webhook_delivery: FAILING (2/3 rules proved)
    RULE-1: FAIL → Fix: test_webhook_basic is failing

Step 4 — Commit proof files (mandatory)

After proof files are written, commit them:

git add specs/**/*.proofs-*.json
git commit -m "test(<feature>): <passed>/<total> rules proved"

For multi-feature runs:

git commit -m "test: run unit tier (<passed>/<total> features fully proved)"

Proof files are project records, not ephemeral build artifacts. Uncommitted proof files make sync_status output inconsistent with what purlin:verify and purlin:drift see.

Writing Tests with Proof Markers

When tests are missing, write them with proof markers. For marker syntax (pytest, Jest, Shell), see references/formats/proofs_format.md. For test quality rules (what makes a proof STRONG vs HOLLOW), see references/audit_criteria.md.

Note

This skill does NOT issue verification receipts. That is purlin:verify's job. This skill runs tests, emits proof files, and reports coverage.