scan-tst

name	scan-tst
description	Use when: analysing a project's test health — coverage, robustness, and CI integration.
argument-hint	Project directory or specific module to assess

Test Perspective (TST)

Evaluates whether the project has a functioning, proportionate test suite. Assigns a score from 0.0 to 1.0 across two dimensions.

Dimensions

Coverage ratio — Are enough source files tested?

• Score = min((testFileCount / sourceFileCount) * 2, 1.0) — 50%+ coverage = full score
• 0 test files = 0.0 immediately (no further analysis needed)

Test script — Can the suite be run?

• package.json must define scripts.test
• Missing test script = 0.0 on this axis regardless of file count

Scoring

tst_score = (coverage_ratio + test_script) / 2

Edge case: if no test files exist at all, score is 0.0 and only one critical task is generated.

When test suite execution fails: If the test suite cannot be executed (missing dependencies, syntax errors, unavailable runtime), record this as a data gap rather than failure; note it in output and do not penalize the coverage score until the execution blocker is resolved.

Score interpretation:

• 0.9–1.0 Test suite is healthy and runnable
• 0.7–0.9 Good coverage but minor gaps
• 0.5–0.7 Coverage is below target — significant untested surface
• 0.0–0.5 Test suite is absent or severely inadequate — highest risk area

Tasks Generated

Finding	Priority
No test files at all	critical
Test file ratio < 50%	high
No scripts.test in package.json	high

Agent Analysis Guide

When performing a manual TST analysis:

1. Count source files (excluding test files) and test files separately
2. Calculate the ratio and identify which source modules have zero test coverage
3. Run the test suite if possible — note any failing tests, not just missing ones
4. Check test speed: tests that take >30 seconds block CI feedback loops
5. Look for edge cases that are clearly missing: empty inputs, null values, concurrent access, error paths
6. Identify tests that only assert the happy path — these create false confidence
7. Flag any test that mocks the database or external service when an integration test is possible

Output: ratio score, list of untested modules ordered by risk, and specific edge cases missing from existing tests.