What problem does it solve? Code reviews and test reviews in the agent-core-v2 engine often lack consistent criteria, producing subjective severity labels and uneven feedback. This Skill supplies calibrated review lenses — single-level-of-abstraction analysis and contract-driven test rules — so findings are reported as factual evidence with file:line references instead of opinionated verdicts. ## Core Features & Use Cases - Abstraction-Level Review (slop lens): Detects functions that mix levels of abstraction or handle errors inline, reporting raw counts, evidence locations, and proposed moves (push down into helpers or up into edge handlers). Invoked only on explicit request. - Test Writing & Review (test lens): Applies contract-driven rules — one behavior per test, drive through the public surface, stub only true external boundaries — for both authoring new tests and auditing existing ones. - Use Case: When reviewing a pull request in packages/agent-core-v2, ask for the abstraction lens on a complex function and receive per-check detections, control-flow counts, exemption flags, and concrete refactoring directions without subjective severity grades. ## Quick Start Review this function in packages/agent-core-v2 using the single level of abstraction lens and report where each foreign concern should move.
