What problem does it solve? Engineering teams often receive vague or noisy code reviews that mix stylistic preferences with real architectural issues, or get pushed toward costly rewrites when a small patch would suffice. This Skill delivers a disciplined, senior-level review of an existing codebase that separates genuine defects from taste, ranks findings by leverage, and recommends the smallest viable change. ## Core Features & Use Cases - Principle-grounded critique: Every recommendation cites a named engineering principle (smallest viable change, distillation, surgical change, short feedback loops, subtraction over addition) drawn from Karpathy, Cherny, and Dohmke. - Structured six-step workflow: Frame, survey, diagnose, prioritize, recommend, and explicitly state what NOT to do, preventing premature-abstraction and rewrite-happy advice. - Standardized report output: Produces a fixed-format report with system read, strengths to preserve, top 3 highest-leverage changes, detailed findings with file:line references, and open questions. - Use Case: Ask it to review a service that has grown organically for two years. It reads the entry points, API surface, tests, and build scripts, then returns a report identifying three high-leverage simplifications (e.g., a single-implementer interface to delete, a slow test loop to fix) plus a list of tempting refactors to avoid. ## Quick Start Ask the agent to review this codebase for over-engineering and tell me the top three highest-leverage changes to make.
