← ClaudeAtlas

github-repo-auditlisted

Audit and rewrite a GitHub repo's README, About sentence, topic tags, and hero asset strategy. Spawns 5 specialist research agents in parallel (competitive landscape, target-user pain language, product positioning, README visual hierarchy + asset hygiene, SEO + AI-citation surface), synthesizes findings, drafts a severity-ranked issue list, gates on user approval, then applies fixes — rewriting README.md, updating GitHub About + topic tags via `gh repo edit`, and producing a hero-asset shotlist (handed to `claude-gif` if installed). Use whenever the user wants to audit, rewrite, reposition, or fix the public-facing surface of any open-source repo, including phrasings like "fix my readme", "rewrite my readme", "audit my github project", "my repo isn't getting stars", "open source positioning help", or pasting a repo URL and asking what's wrong with it. Also triggers on "/github-repo-audit", "repo audit", "readme audit", "repo positioning audit", "open source readme audit". Especially valuable when the repo nam
dancolta/github-repo-audit · ★ 1 · AI & Automation · score 77
Install: claude install-skill dancolta/github-repo-audit
# GitHub Repo Audit Orchestrate a multi-dimensional audit of a GitHub repo's public surface (README, About, topic tags, repo name, hero assets) and apply approved fixes. The audit phase is read-only; the apply phase is gated on explicit user approval per change. ## When this skill fits Use this skill when the user wants to improve how their repo is *perceived* — by humans browsing GitHub, by Google, and by AI engines (Perplexity, ChatGPT, Claude) summarizing the project. It is not a code audit, security audit, or test-coverage audit. It is a positioning + discoverability + first-impression audit. Strong-fit signals: - Repo has a working product but the README reads like commit notes or marketing fluff - Repo name is legacy and no longer matches what the project does - Project competes with a well-known SaaS and needs sharp anti-positioning - Project is human-in-the-loop / responsible-tooling and risks being lumped in with spammy alternatives - Star count or install rate feels low relative to product quality Poor-fit signals (decline and redirect): - User wants code review → suggest `code-reviewer` - User wants security audit → suggest `security-review` or `cybersecurity` - User wants Google SEO for a marketing site, not a repo → suggest `seo` ## Core principles **Read first. Edit second. Push never (without approval).** The 5 research agents run read-only. The synthesis is a draft. Every file write and every `gh` mutation requires explicit user sign-off. **Ground trut