What problem does it solve? AI agents and developers often generate large volumes of Go tests that look productive but detect no real production defects, creating false confidence, slowing CI, and adding maintenance burden. This Skill performs an adversarial audit of new or changed tests to identify which ones fail to justify their existence. ## Core Features & Use Cases - Six-smell classification: Detects self-serving, tautological, linter-replaceable, redundant, phantom-coverage, and integration-obvious tests, each with a DELETE, REWRITE, or MERGE verdict. - Project-aware auditing: Loads the Clawker test infrastructure context (mocks, fakes, testenv, E2E harness) before reviewing, and flags tests that ignore available helpers or bypass the Cobra Factory entrypoint. - Hardcoded-value detection: Traces asserted values back to production code to catch assertions that never exercise real logic, while accepting golden-file and store round-trip patterns. - Use Case: After an agent writes tests for a new CLI command, run the audit to receive a structured report listing each flagged test, its smell category, file location, and a concrete remediation such as replacing a fake-based unit test with an E2E harness test. ## Quick Start Audit the Go tests changed in my current branch against main and report which ones should be deleted, rewritten, or merged.
