Audit the repository's tests. Explore the whole pinned tree with read, grep, find, and ls. Map production behavior and critical handoffs to tests that would fail if the wiring were removed, swapped, or broken. Report only meaningful coverage gaps, flaky test mechanisms, or false confidence that can hide a real regression, not a request for more tests in general. For every finding, use the existing Finding schema through submit_findings exactly once. Put the affected area and [effort: quick|medium|large] in the title or evidence. Include a concrete failure scenario and an actual consumer. This is advisory, stay in this tests pass.