Review quality
Curated catalog of portable Agent Skills for Claude Code, OpenAI Codex, and opencode — an opinionated product→build pipeline (brainstorming, discovery, PRD, research, planning, review) plus skill authoring, security scanning, and infra helpers. Author once, install into any project.
npx -y skills add gabriel-f-santos/coding-agents --skill review-qualityAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.
- 0 stars0 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 changed code for "crap" reduction — reuse, simplification, efficiency, dead code, duplication, naming, and altitude. Use to clean up what was just built — "reduz a gambiarra", "revisa qualidade do código", "tem duplicação/dead code?", "dá pra simplificar essa feature?". Reports findings only (does not apply fixes — use simplify/refactor for that), confidence-based, never security or correctness (those are review-security / review-functionality). Runs standalone or as a subagent of review-phase.
SKILL.md
2.6 KB, as published. Nobody here has run it
review-quality — reduce the crap
Review the changed code for maintainability and waste. Reports opportunities; it does not
apply them (the simplify/refactor skills do that). Quality findings are never P0 and never
about behavior — only how the code reads and runs.
Inputs
The change set (files/diff). Look for the patterns below in the changed code, and at how it fits the surrounding codebase (is it reinventing something that already exists?).
What to look for
- Reuse over reinvention — the change reimplements a helper/util/component that already exists in the repo. (Grep for the existing one; cite it.) The biggest win.
- Duplication — the same logic copy-pasted across the diff (or against existing code) that should be extracted once.
- Dead code — unreachable branches, unused vars/params/imports/exports, commented-out blocks, feature flags never read.
- Over-complexity — a function doing too much; deep nesting; a clever one-liner that hides intent; needless abstraction/indirection (over-engineering for a case that isn't there).
- Altitude / wrong layer — business logic in a controller, SQL in a component, a concern living in the wrong module/boundary.
- Efficiency (non-hot-path) — obvious waste: N+1 that isn't a security/perf P0, repeated recomputation, loading more than needed. (A hot-path perf bug is correctness → that's review-functionality's P1.)
- Naming & clarity — names that mislead or don't match the domain language; a comment explaining what a better name would.
Confidence & output
Report findings worth acting on (skip nitpicks). HIGH = clear, with the concrete improvement;
MEDIUM = worth a look; drop LOW style-only noise. Return:
{dimension: quality, severity, confidence, file:line, issue, evidence, fix (the concrete simplification, and which existing thing to reuse if any)}. Prefer fewer, higher-value
findings over a long list of style nits — the goal is less crap, not more bureaucracy.