← ClaudeAtlas

code-qualitylisted

Review code for idiom, correctness, consistency, testability, justifiability, and code volume. Language-agnostic. Flags real problems only. No style nitpicks without a concrete reason.
s3n-ops/agentonaut · ★ 0 · Code & Development · score 72
Install: claude install-skill s3n-ops/agentonaut
# Code Quality Review Review the current diff or the files named by the user. Produce findings only. Do not fix unless explicitly asked. ## Preparation Determine the scope first: - If the user names specific files, read those files directly - For diff review: `git diff --staged` or `git diff HEAD` - For branch review: `git diff $(git merge-base HEAD main)` ## What to check ### Idiom Use the conventions of the language and ecosystem at hand. What counts as idiomatic is determined by, in order: patterns already established in the codebase, the project's own guidelines (such as `AGENTS.md`, `CLAUDE.md` or `CONTRIBUTING.md`), and the established best practices of the language. Flag deviations only when a better idiomatic alternative concretely exists and the deviation has a real readability or maintainability consequence. ### Correctness - Logic errors: wrong conditions, off-by-one, incorrect operator precedence - Race conditions and TOCTOU (time-of-check to time-of-use) windows - Error paths that silently swallow failures - Incorrect assumptions about external state (process lifetime, filesystem, network) - Tests that pass vacuously (always-true assertions, wrong polarity) - File operations without proper error handling ### Consistency - Same problem, same solution: flag when two patterns solve the same thing differently without reason - Naming: mixed conventions in the same scope - Module and tool choices: mixing levels of abstraction in the same file without justif