skills/montagao/skills/repo-elegance-review

repo-elegance-review

SKILL.md

Repo Elegance Review

Overview

Evaluate a codebase for elegance by scanning structure, naming, boundaries, and repo hygiene, then deliver a concise verdict with evidence and prioritized improvements.

Workflow

1) Calibrate

  • Ask 1-2 quick questions if “pretty/elegant” is undefined: what they value (structure vs. velocity vs. readability), scope (whole repo vs. area), and constraints (stage/team size).
  • Proceed with the default rubric if the user does not respond.

2) Quick Scan (lightweight)

  • Map the top-level layout with ls, rg --files, and find . -maxdepth 2 -type d.
  • Read README.md, primary config (package.json, pyproject.toml, etc.), and .gitignore.
  • Sample a few key directories; avoid full-depth reads unless asked.
  • If it is a git repo, check git status -s for untracked noise.

3) Assess Against the Rubric

Evaluate each category; collect 1-3 concrete examples per issue with file paths.

  • Repo hygiene and clutter (root noise, generated artifacts, ignores not matching reality)
  • Information architecture (clear domains, predictable placement, shallow navigation)
  • Boundaries and layering (separation of concerns, minimal cross-coupling)
  • Naming consistency (directories, modules, types, config alignment)
  • Duplication and drift (parallel folders, overlapping responsibilities)
  • Dependency and config coherence (single source of truth, minimal duplicative tooling)
  • Tests and tooling (test discipline, coverage expectations, docs match behavior)
  • Docs and onboarding (README accuracy, quick start reliability, minimal tribal knowledge)

4) Synthesize and Prioritize

  • Summarize the top reasons the repo feels elegant or messy.
  • Order actions by impact on cognitive load and maintainability.
  • Tie each action to its rationale and the evidence that motivated it.

Output Format

Use this structure:

  • Verdict: Pretty / Mixed / Not pretty
  • Why: 2-3 bullets with the main reasons
  • Strengths: 2-5 bullets
  • Issues: bullets formatted as [Severity] Issue — Evidence (file paths) — Impact
  • Actions: numbered list; each item includes the change, why it helps, and where to start (file paths)
  • Questions: only if critical uncertainty blocks a confident recommendation

Heuristics

  • Favor clarity and predictability over cleverness.
  • Minimize cognitive load: fewer places to look, fewer naming variants.
  • Prefer cohesive modules and explicit boundaries.
  • Keep the root clean; move artifacts to work/, tmp/, or out/ and align .gitignore.
  • Keep docs synchronized with reality; stale docs are anti-elegant.

Pitfalls to Avoid

  • Do not nitpick formatting if it does not affect comprehension.
  • Do not claim issues without evidence from the repo.
  • Do not run heavy commands or full test suites unless the user asks.
Weekly Installs
6
Repository
montagao/skills
GitHub Stars
2
First Seen
Feb 28, 2026
Installed on
gemini-cli6
opencode6
codebuddy6
github-copilot6
codex6
kimi-cli6