Paste a GitHub repository URL to check its health: an explainable score, a checklist of what to fix with exact fixes, a secret scan, tech stack and architecture. Public repositories need no sign-in; connect GitHub (read-only) for private ones.
Private repositories
Public repositories work without any connection.
Read-only access — EditMyStuff cannot modify your repository. Tokens stay on our server, never in your browser.
Try a public repo like or .
GitHub Repo Doctor is a free GitHub repository checker. Paste the URL of any public repository — or connect a private one with read-only access — and get a health report you can act on: a 0–100 score that explains every point, a checklist of problems ranked by severity with the exact fix and file for each, a scan for exposed secrets, and a map of the tech stack, architecture and dependencies.
Every result is computed from the repository itself — its file tree, GitHub metadata and the key files it reads — not guessed by AI. Use it to audit a project before you publish it, to review a dependency you’re about to adopt, or to understand a codebase you’ve just joined.
A repository health check takes under a minute:
The score is a weighted average of six dimensions, each scored 0–100 from signals in the repository and shown with the reasons that moved it: documentation (20%), testing (20%), dependencies (15%), security (15%), maintainability (15%) and project structure (15%). Grades run A (90+), B (80+), C (65+), D (50+) and F.
The checklist is separate and simpler: each check passes, fails or is “not checked”, and the summary counts passed checks per category, so every number maps to a named check you can open.
npm test fail in every CI tool.npm audit or Dependabot.GitHub Repo Doctor is a free GitHub repository checker. Paste a repository URL and it produces a health report: an explainable 0–100 health score, a checklist of problems with fixes, the detected tech stack, a scan for exposed secrets, an architecture map, an import-based dependency graph, a code explorer and analysis of open issues.
Up to about 30 checks, depending on the project: README (and whether it has setup and usage sections), license, contributing guide, changelog, security policy, repository description and topics, issue and pull request templates, .gitignore, .editorconfig, linter and TypeScript strict mode, Node version, lockfiles and unpinned dependencies, CI, tests and the npm test script, documented environment variables, committed .env or key files, and exposed secrets. Checks that don’t apply to the project are skipped.
No — it has read-only access and never writes to your repository. Each problem has a copyable fix, and missing files such as a pull request template or .gitignore come with an “Open prefilled on GitHub” link that opens GitHub’s own new-file editor, where you review and commit the change yourself.
The check couldn’t be decided from the data that was fetched — for example GitHub truncated a very large file tree, a file’s content wasn’t downloaded, or tsconfig.json extends another config. Those checks are shown but never counted as failures.
Yes. Public repositories work instantly with no sign-in. For a private repository you connect GitHub with read-only access (a GitHub App you can revoke anytime); private-repository reads go through a server-side proxy, so no GitHub token is ever exposed to your browser.
No password or personal token. Public repositories use GitHub’s public API directly from your browser. For private repositories you authorize a read-only GitHub App; the credentials stay on the server and the browser never receives a token.
No. The score is deterministic — every point comes from a real signal in the repository (README and license, tests and CI, lockfile and dependency count, exposed-secret findings, activity and contributors, project structure). AI, if used, only explains the results; it never sets the score or the checklist.
Yes. Besides the score and the checklist, the report has an architecture map, a dependency graph built from the imports that resolve to real files, a searchable code explorer and an analysis of open issues, so you can find your way around a codebase you don’t know yet.
It builds relationships from actual imports — ES import/export, CommonJS require, dynamic import, TypeScript and Python imports — resolved against the real file tree. An edge appears only when an import resolves to a file that truly exists; it never guesses relationships from similar filenames.
Public analysis runs in your browser against GitHub’s public API; private analysis runs through a read-only server proxy. The optional AI explanation sends only a small, redacted slice of the repository (a structured summary and a few truncated excerpts) to an AI provider through the site’s AI service (using your own key if you add one in AI settings) — never detected secrets, .env files, or the whole codebase.
No. The deterministic analysis only reports what it actually finds, and the AI is constrained to the real files provided and told never to invent files, functions or facts. Dependency versions are compared with the latest release on npm, but vulnerability checks are not included — verify those with npm audit or GitHub Dependabot.
The tool reads the file tree once and fetches only a bounded set of files, skipping node_modules, build output, binaries and media. The dependency graph shows a focused, explorable subset rather than thousands of nodes at once, and long issue lists are virtualized, so even large repositories stay responsive.