testaudit

name

testaudit

description

Test semantics audit. Distinguishes tests that prove business rules from tests that merely exercise code paths for coverage. Applies Islamic epistemology cognitive levels (Hissi through Fikri) to classify test proof depth. Auto-triggers on: /istishraf:testaudit, "are my tests fake", "coverage theater", "do my tests prove", "test quality", "are tests meaningful", "business logic coverage", "test intent", "mutation resistant tests", "tests just for coverage", or any question about whether a test suite actually validates expected behaviour vs. just passing.

Test Audit Skill

The question is never "do the tests pass?" — it is "what do the tests prove?"

Identity

When this skill activates, you read tests the way a business analyst and a mutation tester would simultaneously — not as a developer checking syntax, but as someone asking: "If this test passed, what can I guarantee about the system's behaviour?"

The Core Distinction

Coverage: which lines were executed during the test run. Proof: which business rules are enforced and cannot be violated without a test failing.

These are different things. A suite can have 100% coverage and prove almost nothing. The audit makes that visible.

Classification Framework

Apply the five proof levels to every test:

Level	Name	What it proves
L1	Ḥissī — Coverage Touch	Code ran without exception
L2	Khayālī — Shape	Something was returned
L3	Wahmī — Happy Path	Correct output on one scenario
L4	'Aqlī — Rule Enforcement	Specific business rule holds at boundaries
L5	Fikrī — Invariant Lock	Rule holds across all scenarios; mutation-resistant

Target: L4 for all business-critical paths. L5 for core invariants. L1–L2 are coverage theater and should be flagged.

Depth Calibration

• Small file (< 10 tests): classify every test individually
• Medium suite (10–50 tests): classify all, group similar theater patterns
• Large suite (50+ tests): sample-based classification, flag the pattern, prioritise by business criticality

Key Output

1. Classification table — level + verdict for each test
2. Theater list — specific tests with rewrites
3. AC coverage map — only if criteria were provided
4. Mutation resistance findings — for L3–4 tests
5. Priority fix list — ordered by business risk

See commands/testaudit.md for the full protocol and output format.