What problem does it solve? Teams often cannot answer how well their code is actually tested: which components have real tests, which tests assert anything meaningful, which are flaky or skipped, and which load-bearing paths have no test at all. This Skill reads a project's tests as a system and produces verifiable, evidence-backed findings about test coverage, quality, and risk. ## Core Features & Use Cases - Test model inventory: Classifies unit, integration, end-to-end, snapshot, property, fuzz, and doc tests with frameworks, runners, and test-to-main LOC ratios per component, including inline tests like Rust #[cfg(test)] modules that path-based tools miss. - Reference-inferred coverage map: Grades each component as well, adequate, thin, or untested based on which tests reference its files, without running a coverage tool or the test suite. - Quality and gap analysis: Detects weak assertions, over-mocking, nondeterminism (sleeps, wall-clock, randomness, network), flaky and skipped tests, and untested hotspots, main loops, persistence, and error paths, synthesized into a testing posture. - Use Case: After running a Sokrates analysis on a large repository, ask how well the payment module is tested; the Skill counts test sites with its script, cross-references hotspot and architecture findings, and reports that the payment writer has no referencing test at any layer. ## Quick Start Ask the AI to run the testing scan on this repository using the existing _sokrates analysis and report which components are thinly tested or untested.
