iw-ai-core-testinglisted

# IW AI Core — Testing Standards **This skill is mandatory reading for any agent or session that writes, reviews, or designs tests for IW AI Core.** It supplements the generic `tests-impl` / `backend-impl` / `tests-review` agents with project-specific rules. - Full strategy: [`docs/IW_AI_Core_Testing_Strategy.md`](../../docs/IW_AI_Core_Testing_Strategy.md) - Test-suite conventions & gotchas: [`tests/CLAUDE.md`](../../tests/CLAUDE.md) - Enhancement roadmap: [`ai-dev/work/TESTS_ENHANCEMENT.md`](../../ai-dev/work/TESTS_ENHANCEMENT.md) - Research basis: [`docs/research/R-00068-ai-core-test-quality-strategy.md`](../../docs/research/R-00068-ai-core-test-quality-strategy.md) --- ## 0. The one rule that matters **Nearly all of IW AI Core's production *and* test code is written by LLM agents.** AI-generated tests frequently *pass while catching nothing*. Before writing each assertion, ask: > **"If I change the production code to return a different value, flip this comparison, or delete this line — will this test fail?"** If the answer is *no*, the assertion is worthless. Strengthen it or delete the test. Every assertion must be one a future mutation-testing run would expect to kill a mutant with. --- ## 1. Assertion strength (NON-NEGOTIABLE) ### Anti-patterns to NEVER write ```python # BAD — tautological; survives every mutation assert result is not None assert isinstance(result, dict) assert len(result) > 0 assert "status" in result assert result # truthiness ``` ```pyth