Gentleman-Programming/gentle-ai v1.15.8 on GitHub

Fix: SDD TDD Agent — Assertion Quality Enforcement

The SDD apply agent was generating trivial tests that gave false confidence:

expect(true).toBe(true) — tautologies that prove nothing
Assertions over empty arrays (queryAllByTestId returning []) inside loops that never iterate
Stopping at GREEN without triangulating (incomplete TDD cycle)

sdd-apply/strict-tdd.md

Added Assertion Quality Rules section with explicit banned patterns: tautologies, orphan empty checks, type-only assertions, ghost loops, incomplete TDD cycles
Defined What Makes a REAL Assertion (3 criteria: calls production code, asserts specific value, breaks when logic changes)
Empty Collection Rule: toEqual([]) only valid with precondition + companion non-empty test
Tightened triangulation: minimum 2 test cases per behavior, skip ONLY for purely structural tasks
Trivial GREEN detection: warns about tests passing because setup doesn't exercise the code path

sdd-verify/strict-tdd-verify.md

Added Step 5f: Assertion Quality Audit — scans test files for all banned patterns
Ghost loops (assertions inside for over empty queryAll results) flagged as CRITICAL
Incomplete TDD cycles (component not rendered due to state) flagged as CRITICAL
Checks triangulation variance — warns if all test cases assert the same trivial value
New Assertion Quality Report Table in verification output