fix-tests-execute

Phase 2: Fix Execution

Invariant Principles

1. Read before fixing — Read the test file and production code before any changes; never guess at code structure.
2. Verify the fix, not just the pass — A test that passes after modification must be confirmed to catch the originally identified blind spot.
3. One fix per commit — Each work item fix is verified and committed independently for traceability and safe rollback.
4. ALL output demands exact equality — See <FORBIDDEN> below and patterns/assertion-quality-standard.md.
5. Fixes must reach Level 4+ — Level 1→2 is still banned. Replacing one weak assertion with another is NOT a fix.

Read patterns/assertion-quality-standard.md in full before writing ANY fix.

The Full Assertion Principle

Every assertion MUST assert exact equality against the COMPLETE expected output — static, dynamic, or partially dynamic. For dynamic output, construct the complete expected value dynamically, then assert ==.

# CORRECT
assert result == "the complete expected output, every character"

# CORRECT - dynamic output
assert message == f"Today's date is {datetime.date.today().isoformat()}"

assert "substring" in result is BANNED. ALWAYS. NO EXCEPTIONS. Dynamic content is no excuse for a partial check.

When fixing a partial assertion on dynamic output, construct the complete expected value using the same logic the function uses, then assert ==. Prefer construct-then-compare over normalization; normalization is a last resort for truly unknowable values only (random UUIDs, OS-assigned PIDs, memory addresses).

When fixing a partial mock assertion, also check whether ALL mock calls are fully asserted. Assert EVERY call with ALL args and verify the call count. NEVER use mock.ANY — construct the expected argument dynamically when it is dynamic.