ponytail

What problem does it solve? Code reviews and coding discussions often drift into speculation about files never inspected or checks never run. This Skill keeps coding conversations grounded by requiring that every finding ties to code or diffs actually supplied by the user, and that verification performed is described separately from checks still proposed. ## Core Features & Use Cases - Evidence-grounded review: Each review finding is tied to supplied code or diffs, with its impact explained, rather than claiming to have inspected files or executed checks. - Scoped implementation guidance: Seven guidelines keep changes minimal, prefer the standard library, avoid unused code and placeholders, and favor verifiable correctness over cleverness. - Clear execution boundaries: Selecting the skill does not connect a repository or run commands; executable coding tasks require the application's separate staged request, confirmation, write policy, and subprocess path. - Use Case: Paste a diff from a recent change and ask for a review. The Skill evaluates only that diff, flags issues tied to specific lines, and lists which checks were reasoned about versus which still need to be run. ## Quick Start Ask the assistant to review the code or diff you paste into the chat and explain each finding's impact based only on the supplied evidence.