Snapshot · 2026-08-31 · exact commits

How ready are public repositories for coding agents?

Name: Repo Agent Kit Agent Readiness Benchmark
Creator: Repo Agent Kit

A transparent snapshot of 10 public development repositories across agent guidance, reproducible setup, visible tests, CI, and contributor guardrails. Every score links to its pinned source evidence.

Download JSON

Average score

4/10

Root AGENTS.md

2/10

Agent-ready

Pinned commits

Commit-pinned agent readiness scores for ten public repositories
Repository	Score	Status	Agent guidance	Evidence	Re-run
1 vercel/next.js canary · d434afa837b9	96/100	Agent-ready	AGENTS.md File quality 82/100	readme.md package.json pnpm-lock.yaml jest.config.js	Live audit
2 rust-lang/rust main · 0dfb098f3aee	94/100	Agent-ready	AGENTS.md File quality 70/100	README.md Cargo.toml Cargo.lock tests	Live audit
3 microsoft/vscode main · bcec3677a1c1	85/100	Good foundation	AGENTS.md File quality 26/100	README.md package.json package-lock.json test	Live audit
4 openai/codex main · 13bc770eaf0a	83/100	Good foundation	AGENTS.md File quality 54/100	README.md package.json pnpm-lock.yaml .github/workflows	Live audit
5 vitejs/vite main · 238ad811c7fb	62/100	Needs guidance	.github/copilot-instructions.md	README.md package.json pnpm-lock.yaml vitest.config.e2e.ts	Live audit
6 django/django main · 189136c2a3e1	54/100	Needs guidance	.github/copilot-instructions.md	README.rst pyproject.toml tests .github/workflows	Live audit
7 facebook/react main · 065bc84eb509	54/100	Needs guidance	CLAUDE.md	README.md package.json yarn.lock .github/workflows	Live audit
8 tiangolo/fastapi master · 49033471594e	43/100	Not agent-ready	None detected	pyproject.toml uv.lock tests .github/workflows	Live audit
9 golang/go master · f534300f9830	25/100	Not agent-ready	None detected	test CONTRIBUTING.md	Live audit
10 anthropics/claude-code main · f275fa282e76	24/100	Not agent-ready	None detected	.github/workflows SECURITY.md	Live audit

Scores are comparative evidence snapshots, not judgments about overall code quality or maintainers. A missing signal can mean the evidence lives deeper than the audit scope.

Reproducible methodology

Eight checks, no model judgment.

The snapshot reads the repository root plus direct children of .github and .cursor. It uses the same deterministic rubric as the live audit.

Every row is pinned to an exact Git commit.

Evidence links point to files at that commit, not a moving branch.

No repository code is executed and no private data is requested.

Root AGENTS.md or tool-specific guidance

30 pts

AGENTS.md instruction quality

20 pts

Root README

10 pts

Build manifest

10 pts

Dependency lockfile

8 pts

Visible root test surface

8 pts

Recognized CI surface

7 pts

Contribution, security, or ownership guidance

7 pts

Scope limit: This static audit measures discoverable repository evidence. It does not execute code, inspect nested package quality, or predict task success.

Check a different repository

Run the same rubric live in your browser.

Paste any public GitHub repository. No sign-in or GitHub token required.

Audit a repository