GitHub Repo AI Readiness Score

Score any public GitHub repo on AI-readiness. 0-100 + AI-Native / AI-Friendly / AI-Curious / AI-Aware / AI-Absent tier + per-signal breakdown. Stop guessing whether a repo plays nice with Claude Code, Cursor, Copilot, Aider, Continue.dev, and Windsurf.

What does GitHub Repo AI Readiness Score do?

In 2026 every popular AI coding agent (Claude Code, Cursor, GitHub Copilot, Aider, Continue.dev, Windsurf, Cline) reads a slightly different set of project-config files: CLAUDE.md, AGENTS.md, .cursorrules, .github/copilot-instructions.md, CONVENTIONS.md, .continuerc.json, .windsurfrules, and so on. The presence or absence of these files determines whether agents work brilliantly or burn tokens guessing.

This Actor inspects the default branch of a public GitHub repo, audits which signals are present, and outputs:

• An AI-readiness score from 0-100
• A tier label (AI-Native >= 80, AI-Friendly >= 60, AI-Curious >= 40, AI-Aware >= 20, AI-Absent < 20)
• The list of high-impact missing files so you know what to add
• A full per-signal breakdown with weights

It is the repo-level companion to gh-org-ai-fingerprinter (#9), which scores at the org level. Use this Actor when you want one repo at a time, or batch it with a list of owner/name inputs.

Why use GitHub Repo AI Readiness Score?

• Picking dependencies: an AI-Native library will give your agent better context than an AI-Absent one — favor those for vibe-coding sessions
• Onboarding a repo to Claude Code / Cursor: see what's missing before you hit the first error
• OSS maintainers: identify gaps in your AI-readiness and add files to reduce drive-by AI-generated PR noise
• Recruiters / scouts: an AI-Native engineering culture is a strong hiring signal in 2026
• Investors: portfolio-wide AI-readiness audit of every dep tree of every portfolio company

How to use GitHub Repo AI Readiness Score

1. Enter a public GitHub repo as owner/name or full URL (e.g., anthropics/claude-code or https://github.com/microsoft/playwright)
2. Click Start
3. Open the dataset row OR download ai-readiness-report.md from the Storage tab

Input

• githubRepo — the repo to score, in owner/name form or full GitHub URL

Output

You can download the dataset in various formats such as JSON, HTML, CSV, or Excel. Each row contains:

{
  "repo": "anthropics/claude-code",
  "url": "https://github.com/anthropics/claude-code",
  "defaultBranch": "main",
  "stars": 8420,
  "description": "Official CLI for Claude Code",
  "language": "TypeScript",
  "aiReadinessScore": 84,
  "tier": "AI-Native",
  "detected": {
    "claudeMd": true,
    "agentsMd": false,
    "cursorRules": true,
    "copilotInstructions": false,
    "claudeDir": true,
    "claudeAgents": true,
    "claudeSkills": true,
    "claudeHooks": true,
    "claudeSettings": true,
    "aiderConventions": false,
    "continueRc": false,
    "windsurfConfig": false,
    "mcpConfig": true,
    "githubPromptIssueTemplates": true,
    "ciAi": true,
    "readmeMentions": true,
    "description": true
  },
  "missingHighValue": ["AGENTS.md at root", "GitHub Copilot instructions"]
}

Data table

Field	Meaning
repo	owner/name path
stars	Star count
aiReadinessScore	0-100 composite
tier	One of: AI-Native, AI-Friendly, AI-Curious, AI-Aware, AI-Absent
detected	Per-signal boolean map
missingHighValue	List of weight >= 6 signals not present
signals	Full breakdown with weight + present flag

Signal weights

Signal	Weight
CLAUDE.md at root	18
AGENTS.md at root	8
.cursorrules / .cursor/rules	8
GitHub Copilot instructions	7
.claude/ directory	6
.claude/agents/	6
.claude/skills/	6
.claude/hooks/	6
.claude/settings.json	5
MCP config (mcp.json)	6
Aider CONVENTIONS.md	3
Continue.dev (.continuerc.json)	3
Windsurf (.windsurfrules)	3
README mentions Claude/Cursor/Copilot	4
CI workflow tagged AI/agent	3
Issue templates present	2
Repo description mentions AI/LLM	2

Pricing / Cost estimation

Pay-per-event:

• First scan: free trial
• Per-repo scan: $0.05
• Bulk pack (50 repos): $2
• Monthly portfolio audit subscription: $4/month

Each scan uses only 2 GitHub API calls (repo metadata + git tree), so even free anonymous quotas (60/h) support ~30 scans per hour.

Tips and Advanced options

• Set GITHUB_TOKEN in the Actor's Source > Environment Variables tab for a 5000 req/h ceiling (lets you scan thousands of repos in one run)
• Combine with the org-level companion gh-org-ai-fingerprinter (#9) for a top-down audit
• Pair with claude-md-completeness-grader — the readiness score tells you if a CLAUDE.md exists; the grader tells you how good it is

FAQ, disclaimers, and support

What is "AI-readiness"?

It's a measure of how much project-specific context an AI coding agent can read on first sight. A repo with CLAUDE.md + AGENTS.md + .cursorrules + a populated .claude/ directory will produce dramatically better agent output than a repo with nothing — the agent doesn't have to guess at conventions, security rules, or testing setup.

Why is the weight for CLAUDE.md so high (18)?

CLAUDE.md is the broadest and most-actioned file. Claude Code reads it on every session, and other agents (Cursor, Cline) have started honoring it too. It's the single highest leverage signal.

Limitations

• Only inspects the default branch — content on feature branches is ignored
• Detection is file-existence based; a .cursorrules file that's empty still counts (use the companion grader to check quality)
• For very large monorepos (>100k files) the GitHub git-tree response truncates and some deep paths may be missed; a warning is logged

Legality

This Actor only reads metadata and file listings via the public GitHub REST API. No private repos, no scraping.

Support

Companion repo: github.com/ianymu/claude-verify-before-stop — a Claude Code Stop hook that catches "task complete" hallucinations before they ship.

Built by Ian Mu — github.com/ianymu. Open an issue on the Actor's Issues tab if you find a missing AI-tool signal worth tracking.