Code quality review
Skill batteryshark/skill-tap/skills/development/code-quality-review
Portable, reusable skills for agents and everyday workflows. Clank responsibly.
npx -y skills add batteryshark/skill-tap --skill code-quality-reviewAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 26 days oldThe repository was created 26 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Review or improve source code for reader-facing maintainability across naming, functions, models, comments, error handling, tests, external boundaries, and concurrency. Use for broad code-quality audits, behavior-preserving cleanup, generated-code humanization, refactoring prioritization, or focused requests to reduce local cognitive load without redesigning the whole architecture.
SKILL.md
3.4 KB, as published. Nobody here has run it
Code-quality review
Improve code from repository evidence rather than a universal style checklist. Optimize for accurate intent, safe change, and the conventions of the language and project under review.
Choose a mode
Use audit mode when asked to review, diagnose, or prioritize. Report findings without editing.
Use cleanup mode when asked to improve or refactor. Make small, behavior-preserving changes and validate each coherent step.
Workflow
- Read manifests, source layout, formatter and linter configuration, tests, and representative execution paths.
- Run
bin/code-quality-review <repo>for a neutral inventory of languages, large files, tests, and debt markers. It follows Git ignore rules by default; use--include-ignoredonly when ignored source is explicitly in scope. - Classify concrete issues with
references/review-rubric.md. Ignore generated, vendored, dependency, build, and minified code unless it is explicitly in scope. - When needless machinery is the concern, use
references/complexity-pruning.mdto look for deletion and consolidation opportunities without assuming every abstraction is waste. - Prioritize correctness hazards, hidden failure behavior, missing behavioral coverage, and high-cost comprehension problems before cosmetic consistency.
- In cleanup mode, follow
references/safe-refactoring.md. Preserve public APIs, serialized shapes, persisted data, and externally observable behavior unless migration is explicit. - Apply the repository's language conventions. Read
references/language-guidance.mdonly for the languages present. - Run the narrowest relevant tests, type checks, linters, formatters, or build commands.
Rules
- Treat metrics and patterns as investigation leads, not verdicts.
- Explain the reader or behavior cost of every finding.
- Prefer domain vocabulary over generic names, but do not rename public contracts casually.
- Split functions, classes, or modules by coherent responsibility rather than line-count thresholds.
- Remove duplication only when the shared concept is real.
- Keep comments for intent, constraints, hazards, and external quirks; improve code when a comment would only narrate it.
- Prefer deleting dead flexibility, using an existing platform capability, or consolidating a pass-through layer before introducing another abstraction.
- Prefer tests of observable behavior over tests that freeze private structure.
- Do not infer poor quality or authorship merely because code looks generated.
Report
For audits, order findings by risk and include evidence, impact, action, and confidence. For edits, summarize reader-facing improvements, protected behavior, checks run, and residual risk.
Use agents/auditor.md for an independent review, agents/simplifier.md for a deletion-focused pass, and agents/fixer.md for a bounded cleanup pass.