What problem does it solve? New karenina adapters can pass cold tests while hiding bugs that only appear with live API calls, and they can drift from project conventions. This Skill provides a structured final review phase that catches quality, correctness, and convention issues before an adapter is merged. ## Core Features & Use Cases - Convention and code quality checks: Verifies file size limits, lazy import patterns, __all__ exports, no cross-adapter imports, and CLAUDE.md style rules such as PEP 604 unions and Google-style docstrings. - Hot-test-proven correctness checks: Validates structured output JSON serialization, max_turns wiring, timeout handling on all execution paths, MCP session lifecycle, and partial-result recovery on timeouts. - Registration and regression verification: Confirms AdapterRegistry registration, plugin entry points, full test suite passage, and a minimum hot test set (H1, H2, H5, H6, H8, H9) with live API calls. - Use Case: After implementing and testing a new LangGraph adapter for karenina, run this review to verify it honors timeouts, produces valid structured output, and complies with all project conventions before merging. ## Quick Start Review the new adapter at karenina/src/karenina/adapters/my-adapter against the review checklist and confirm it is ready to merge.
