Audit and configure the public GitHub repository surface: launch readiness, repo quality/adoption/trust, discoverability, contribution readiness, topics, homepage URL, description, issues/wiki/license/CI settings, owner/org-wide repo audits, repo popularity/adoption signals, README/package/GitHub metadata drift, URL/license/topics gaps, and Agent Skill repo packaging/evals. Use when the user asks for repo-level readiness, repo metadata/configuration, or visitor/contributor trust. Do not trigger for README-only writing, one PR descriptions/reviews, function-level code tests, or general implementation/architecture work; defer those to narrower specialists unless the user explicitly asks for repo-level readiness/proof/metadata judgment.
Installer avec Codex ou Claude Copiez ce prompt, collez-le dans Codex, Claude ou un autre assistant, puis laissez-le vérifier la page du skill et l'installer pour vous.
Une commande directe contourne le prompt de vérification. Examinez la source avant de l'exécuter.
Audit and configure the public GitHub repository surface: launch readiness, repo quality/adoption/trust, discoverability, contribution readiness, topics, homepage URL, description, issues/wiki/license/CI settings, owner/org-wide repo audits, repo popularity/adoption signals, README/package/GitHub metadata drift, URL/license/topics gaps, and Agent Skill repo packaging/evals. Use when the user asks for repo-level readiness, repo metadata/configuration, or visitor/contributor trust. Do not trigger for README-only writing, one PR descriptions/reviews, function-level code tests, or general implementation/architecture work; defer those to narrower specialists unless the user explicitly asks for repo-level readiness/proof/metadata judgment.
license
MIT
compatibility
Agent Skills clients including Codex, OpenCode, Pi, Gemini CLI, and Claude Code.
good-repo is a repository effectiveness skill. It treats a GitHub repo as a product surface that must help visitors answer:
What is this?
Why should I care?
Can I trust it?
Can I try it quickly?
Can I contribute or maintain it?
Quality and success feature model
A repo's quality is not measured by stars first. Stars are an outcome. Audit the features that make the right visitor understand, trust, try, use, and contribute:
Clear front door — GitHub description + README quickly explain what, who, why, and how to try.
Proof that it works — screenshots, demo, terminal output, runnable examples, evals, tests, notebooks, or sample responses fit the project class.
Fast adoption path — prerequisites, install, quick start, expected output, local dev path, package metadata.
Accurate documentation — README and docs match current code, links work, examples are real.
Focus and positioning — clear scope, non-goals, audience, and when not to use it.
When auditing, always include GitHub's own baseline surfaces: README, Community Profile files, description, homepage, topics, license, contribution guidance, security policy when relevant, CI/workflows, and large-file/repo-size risks. Load references/github-official-baseline.md when the user asks what GitHub itself recommends or when a finding should be grounded in official GitHub docs. Load references/community-profile-repolinter.md when using GitHub Community Profile, GitHub's community health percentage/API, or Repolinter-style policy checks.
When the user asks about repo popularity, stars, forks, growth, or adoption impact, load . Treat stars/forks/watchers as noisy outcomes, not intrinsic quality. State correlations carefully and avoid claiming that one repo hygiene change will cause popularity.
Always include the homepage URL configuration check: if the README, manifest, or docs expose a likely demo/docs/homepage URL, verify GitHub's repository homepageUrl is set to that URL. If it is empty or points elsewhere, flag it as a discoverability/configuration issue.
The reference exemplars are:
nutlope/hallmark — productized skill repo with clear front door, proof gallery, live demo, install path, packaged skill, modular references, generation tests, roadmap, and license. Load references/hallmark-exemplar.md when explaining or applying the Hallmark pattern.
adewale/good-pr — small operational skill repo with real-pain origin, trigger-rich skill description, PR templates, self-review checklist, readiness script, and evals. Load references/good-pr-exemplar.md when explaining or applying the good-pr pattern.
This skill complements good-readme. When README creation/improvement is substantial, use good-readme if available; otherwise apply the README gate embedded in this skill.
Verbs
Invocation
What it does
(default)
Audit current repo, infer project class, score effectiveness, and propose fixes.
good-repo audit [target]
Read-only audit. Target can be current repo, local path, or public GitHub URL. No edits.
good-repo configure
Apply safe, non-destructive repo improvements after showing the file/settings plan.
good-repo launch
Prepare repo for public launch: README handoff, metadata, license, proof, quick start, examples, CI, contribution flow.
Explain why a repo is exemplary or weak, with transferable patterns.
good-repo owner-audit <owner>
Assess every public repo under a GitHub owner/profile; score, detect homepage drift, and summarize portfolio cleanup themes.
If the user asks generally to "make this repo better," run the default audit first, then ask before editing. If they ask to assess a GitHub owner/profile, use skills/good-repo/scripts/audit-github-owner.py <owner> when available, then manually review the high-value repos and any homepage URL drift candidates.
Trigger policy
Strong triggers — load this skill
Use good-repo when the user asks about any of these:
Repo quality, success, or popularity — "is this repo good?", "make this repo exemplary", "why isn't this repo getting adopted?", "will repo quality affect stars?", "repo effectiveness", "launch-ready".
GitHub launch/configuration — topics, description, homepage URL, social preview, Community Profile health, issues/discussions/wiki, license detection, branch protection, releases, templates.
Repo audits — current repo, public GitHub URL, OSS project, owner/org-wide assessment, random repo sample, score/recommendation table.
Metadata drift — README vs package vs GitHub mismatch; live/demo/docs URL not configured as GitHub homepage; README license vs missing root LICENSE.
Adoption/trust gaps — no quick start, no proof, no examples, stale screenshots, missing CI, missing changelog, missing contribution path.
Exemplar comparison — Hallmark, good-pr, good-readme, ripgrep, Flask, or any repo used as a model.
Soft triggers — load if repo-level judgment is needed
README improvement that also asks about launch, metadata, proof, examples, or trust.
PR/contribution process at the repository level, not drafting one PR.
CI/release/dependency automation when the question is about public repo readiness.
Tests/evals as repo proof, especially for Agent Skill repos.
Do not trigger / defer to specialists
README-only writing or rewriting → prefer good-readme; return after it for repo-level gaps.
One PR description or pre-submit PR review → prefer good-pr.
Writing or improving code tests → prefer testing-best-practices; use good-repo only to assess tests as public proof/CI signals.
General code review, architecture, or implementation with no repo-public-surface question → do not trigger.
GitHub issue triage or project management unless tied to repository quality/configuration.
Trigger decision rule
If the user's real question is "will a visitor/user/contributor trust, try, adopt, or contribute to this repo?" load good-repo. If the question is only about a document, PR, test, or code change, prefer the narrower skill and use good-repo only for the repo-level wrapper.
Safety rails
Before editing, state the exact files and GitHub settings you plan to change. Ask for confirmation before any of these:
Deleting files, rewriting README sections wholesale, or changing public install commands.
Claiming metrics, users, testimonials, benchmarks, compatibility, or security status not evidenced in the repo.
Always distinguish observed evidence from recommended additions. Do not invent stars, logos, adoption claims, screenshots, roadmap commitments, or support channels.
Default workflow
0. Recon
Read enough repo context before judging:
README*, package.json, pyproject.toml, Cargo.toml, go.mod, or equivalent manifest.
For GitHub remotes, inspect metadata if gh is authenticated or the public API is available, including description, homepageUrl, repositoryTopics, license detection, feature toggles, and default branch.
Classify the repo:
Skill / agent package
Library / package
CLI / TUI tool
Web app / SaaS / template
Docs / knowledge repo
Research / ML artifact
Personal / portfolio / experiment
Internal / private operational repo
State the class because "excellent" differs by class.
0a. Proportionality gate
Before applying any public-launch or Community Profile checklist, name the repo's lifecycle and audience. For personal experiments, internal/private repos, forks/reference repos, and historical artifacts, recommendations must be proportionate and right-sized rather than governance theater.
For a tiny personal experiment, prioritize only the lightweight basics that fit the evidence:
README purpose/status, install/run command, and expected result.
Dependency, secret, privacy, or data-loss footguns if present.
License decision only when the repo is public/reusable or the owner asks.
One small smoke check only if the repo is actively maintained or claims repeatable behavior.
Explicitly defer heavy process until the condition exists: external contributors, security-sensitive code/service, package consumers, active releases, or maintainer capacity. Do not recommend CODE_OF_CONDUCT.md, SECURITY.md, issue forms, Dependabot, branch protection, release automation, or a large CI suite just to improve a score. If those surfaces are not applicable, mark them N/A or low-priority with the condition that would make them relevant.
1. README handoff
If README work is a major part of the task:
Prefer loading/running good-readme if it is installed or available in the project.
Use evidence-backed scoring. If a criterion is not applicable, award credit only when the repo's class makes it genuinely irrelevant; otherwise mark as missing.
3. Apply exemplar patterns where useful
Load exemplar references when relevant:
references/hallmark-exemplar.md — when the user asks why Hallmark works, when designing a productized skill/package repo, or when proof/gallery/live-demo discipline matters.
references/good-pr-exemplar.md — when designing a small operational skill, checklist/template workflow, or skill with scripts/evals.
references/skill-repo-best-practices.md — when auditing or configuring an Agent Skill repo's folder structure, package metadata, marketplace metadata, references, scripts, or evals.
Generalize the Hallmark moves:
README is a concise landing page, not the whole manual.
Visual or executable proof appears in the first screenful.
The public API is named clearly (Hallmark's four verbs).
Detailed rules live in modular references, not a giant README.
Examples/recipes show real outputs for different scenarios.
Package metadata makes installation and discovery unambiguous.
A roadmap signals active judgment, not vague TODOs.
Generalize the good-pr moves:
Start from real user/maintainer pain.
Invert complaints into a concrete checklist.
Provide both a blank template and a filled example.
Add a quick self-review checklist for the workflow.
Automate cheap mechanical checks with a small script.
Add evals that test the non-obvious behavior the skill must catch.
Keep the repo small when the workflow is small.
4. Produce prioritized fixes
Separate findings into:
Launch blockers — missing license, no install/run path, broken README commands, no proof for visual/CLI tools, package not installable.
