Skip to content

EShener/agent-ready

agent-ready

Make any repository ready for AI coding agents in 60 seconds.

CI npm version npm downloads agent-ready license

Showcase · Examples · Detectors · Benchmark · Publishing · Growth playbook · Launch kit · Launch notes

agent-ready is a zero-dependency CLI that scans a codebase, generates canonical AI agent instructions, checks them for drift and missing verification steps, and gives the repository an explainable Agent Readiness Score.

It is for developers using Codex, Claude Code, Cursor, Gemini CLI, GitHub Copilot, or any coding agent that needs repository instructions before editing safely.

npx --yes @eshen_fox_mie/agent-ready improve --preset team --dry-run

Preview exactly which files will be added and the estimated score gain before anything is written.

If this saves you setup time, star the repo so other maintainers can find it.

agent-ready posts a pull request comment with score, compatibility, and top fixes

Why This Exists

AI coding agents fail for boring reasons: they do not know the test command, they miss local safety rules, they read stale instructions, or every tool has its own duplicated guidance.

agent-ready turns that tribal knowledge into a small, lintable repository contract.

Benchmark Snapshot

On a 2026-05-23 sample of six public AI/devtool repositories, the average Agent Readiness Score was 28/100. See BENCHMARK.md for the reproducible leaderboard and sample commits.

60 Second Demo

Use the GitHub Action immediately:

- uses: EShener/agent-ready@v0.1.25
  with:
    fail-under: 80

Use the CLI from npm:

npx --yes @eshen_fox_mie/agent-ready doctor
npx --yes @eshen_fox_mie/agent-ready examples
npx --yes @eshen_fox_mie/agent-ready explain
npx --yes @eshen_fox_mie/agent-ready matrix
npx --yes @eshen_fox_mie/agent-ready comment
npx --yes @eshen_fox_mie/agent-ready compare --before before.json --after after.json
npx --yes @eshen_fox_mie/agent-ready improve --preset team --dry-run
npx --yes @eshen_fox_mie/agent-ready improve --dry-run
npx --yes @eshen_fox_mie/agent-ready improve --dry-run --format issue
npx --yes @eshen_fox_mie/agent-ready improve --level team
npx --yes @eshen_fox_mie/agent-ready fix --preset team --dry-run
npx --yes @eshen_fox_mie/agent-ready fix --dry-run
npx --yes @eshen_fox_mie/agent-ready fix --level team --dry-run
npx --yes @eshen_fox_mie/agent-ready init --preset oss --dry-run
npx --yes @eshen_fox_mie/agent-ready init --dry-run
npx --yes @eshen_fox_mie/agent-ready score --fail-under 80
npx --yes @eshen_fox_mie/agent-ready ci
npx --yes @eshen_fox_mie/agent-ready ci --comment
npx --yes @eshen_fox_mie/agent-ready ci --write --dry-run
npx --yes @eshen_fox_mie/agent-ready snapshot --write --dry-run
npx --yes @eshen_fox_mie/agent-ready benchmark ../repo-a ../repo-b
npx --yes @eshen_fox_mie/agent-ready leaderboard ../repo-a ../repo-b
npx --yes @eshen_fox_mie/agent-ready roadmap ../repo-a ../repo-b

Example output:

agent-ready doctor: my-app
Score: 72/100 (C)
Primary language: TypeScript
Commands: install=npm install; test=npm run test
Agent docs: none

Top fixes:
- [warning] Missing AGENTS.md canonical agent instructions.
  Run `agent-ready init --targets codex`.

After review, generate the files:

npx --yes @eshen_fox_mie/agent-ready init --targets codex,claude,cursor,gemini,copilot

What It Generates

  • AGENTS.md as the canonical source of truth
  • CLAUDE.md
  • .cursor/rules/agent-ready.mdc
  • GEMINI.md
  • .github/copilot-instructions.md

The tool-specific files are intentionally small shims that point back to AGENTS.md, so instructions do not drift across tools.

Presets

Use presets when you want an opinionated starter instead of assembling flags by hand.

Preset Best for Adds
oss Public projects and contributor-friendly repos All agent docs plus team collaboration starters
team Internal teams that want CI feedback on PRs oss plus the Agent Ready workflow with PR comments
enterprise Larger repos that need governance defaults team plus .env.example, CODEOWNERS, issue templates, and security policy
agent-ready improve --preset team --dry-run
agent-ready fix --preset enterprise
agent-ready init --preset oss --dry-run

Commands

doctor

Runs scan, lint, and score together. This is the best first command and the best screenshot for issues.

agent-ready doctor
agent-ready doctor --format json
agent-ready doctor --fail-under 80

scan

Detects languages, frameworks, package manager, CI, standard commands, existing agent docs, README, and top-level structure.

agent-ready scan
agent-ready scan --format json

init

Plans or writes agent instruction files. Existing files are skipped unless --force is provided.

agent-ready init --dry-run
agent-ready init --preset oss --dry-run
agent-ready init --preset enterprise --dry-run
agent-ready init --targets codex,claude,cursor
agent-ready init --interactive
agent-ready init --force

fix

Applies the safe generated fixes: agent instruction files plus the Agent Ready CI workflow. Existing files are skipped unless --force is provided.

Use --level team to add collaboration starters such as PR templates, contributing docs, architecture docs, and an ADR. Use --level full to also add .env.example, CODEOWNERS, an agent-readiness issue template, and a security policy.

agent-ready fix --dry-run
agent-ready fix
agent-ready fix --preset team --dry-run
agent-ready fix --preset enterprise
agent-ready fix --level team
agent-ready fix --level full --dry-run
agent-ready fix --targets codex,cursor --no-ci
agent-ready fix --force

improve

Runs scan, applies staged fixes, rescans, and prints a before/after improvement report. Use --dry-run to turn it into a safe launch-post screenshot that shows planned files without writing them. Use --format issue to create a GitHub-ready checklist for contributors.

agent-ready improve --dry-run
agent-ready improve
agent-ready improve --preset team --dry-run
agent-ready improve --preset enterprise --dry-run --format issue
agent-ready improve --level team
agent-ready improve --level full --comment
agent-ready improve --dry-run --format issue
agent-ready improve --format json

lint

Finds missing or risky agent readiness gaps:

  • Missing canonical AGENTS.md
  • Missing test/lint commands
  • Missing README or CI
  • Agent instructions with no safety or verification section
  • Stale path references in agent docs
  • Drift-prone duplicate agent docs
agent-ready lint
agent-ready lint --format json

annotations

Prints GitHub Actions workflow commands for readiness findings, so CI can show warnings and notices inline.

agent-ready annotations
agent-ready annotations --format json

score

Prints a 0-100 Agent Readiness Score with explainable deductions.

agent-ready score
agent-ready score --fail-under 80

Use --fail-under in CI to fail a build when repository instructions regress.

examples

Prints the example gallery with commands and links to copy-ready output samples. The gallery includes real-world repository shapes such as Next.js apps, Docker Compose services, monorepos, and CLI/devtools.

agent-ready examples
agent-ready examples --format json

explain

Prints an impact-ranked fix plan with why each issue matters and how many points each fix can recover.

agent-ready explain
agent-ready explain --format json

matrix

Prints a compatibility matrix for Codex, Cursor, GitHub Copilot, Claude Code, and Gemini CLI.

agent-ready matrix
agent-ready matrix --format json

comment

Prints a concise GitHub-ready Markdown summary for PR comments, issues, discussions, and launch posts.

agent-ready comment
agent-ready comment --max-fixes 5
agent-ready comment --format json

compare

Compares two readiness JSON files and prints a before/after score summary for PR comments, issues, and launch posts.

agent-ready report --format json > before.json
agent-ready fix
agent-ready report --format json > after.json
agent-ready compare --before before.json --after after.json
agent-ready compare --before before.json --after after.json --format json

report

Creates a Markdown or JSON report for issues, PRs, or README updates.

agent-ready report --format markdown

snapshot

Creates a complete static AGENT_READINESS.md report with score, commands, agent compatibility, findings, and next actions.

agent-ready snapshot
agent-ready snapshot --write
agent-ready snapshot --write --dry-run
agent-ready snapshot --write --force
agent-ready snapshot --output docs/agent-readiness.md --write
agent-ready snapshot --format json

badge

Prints a Shields-compatible static badge for README files.

agent-ready badge
agent-ready badge --format url
agent-ready badge --format json
agent-ready badge --fail-under 80

benchmark

Scores multiple local repositories and prints a shareable Markdown leaderboard.

agent-ready benchmark ../repo-a ../repo-b
agent-ready benchmark --root ../work repo-a repo-b
agent-ready benchmark --format json

leaderboard

Scores multiple repositories and prints a share-ready leaderboard with the most common readiness gaps and a launch snippet.

agent-ready leaderboard ../repo-a ../repo-b
agent-ready leaderboard --root ../work repo-a repo-b
agent-ready leaderboard --format json

roadmap

Turns a multi-repository scan into a phased readiness cleanup roadmap for team planning.

agent-ready roadmap ../repo-a ../repo-b
agent-ready roadmap --root ../work repo-a repo-b
agent-ready roadmap --format json

ci

Prints a ready-to-paste GitHub Actions workflow.

agent-ready ci
agent-ready ci --write
agent-ready ci --write --dry-run
agent-ready ci --write --force
agent-ready ci --fail-under 90
agent-ready ci --mode npx

Configuration

Add agent-ready.json when detected commands or docs need overrides.

{
  "targets": ["codex", "claude", "cursor"],
  "docs": {
    "architecture": "docs/architecture.md",
    "decisions": "docs/adr"
  },
  "commands": {
    "install": "npm install",
    "test": "npm run test:ci",
    "lint": "npm run lint"
  }
}

agent-ready init uses configured targets when --targets is omitted. Use --config path/to/file.json to load a non-default config file.

CI Gate

Use the reusable GitHub Action as a lightweight quality gate:

name: Agent Ready

on:
  pull_request:
  push:
    branches: [main]

jobs:
  agent-ready:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: EShener/agent-ready@v0.1.25
        with:
          fail-under: 80

The action writes a diagnosis, compatibility matrix, and Markdown report to the GitHub Actions Step Summary, annotates readiness findings in the Actions UI, then fails the job when the score is below fail-under.

To post or update a pull request comment, grant comment permissions and set comment: true:

permissions:
  contents: read
  pull-requests: write
  issues: write

steps:
  - uses: actions/checkout@v4
  - uses: EShener/agent-ready@v0.1.25
    with:
      fail-under: 80
      comment: true

Generate this workflow with:

npx --yes @eshen_fox_mie/agent-ready ci > .github/workflows/agent-ready.yml
npx --yes @eshen_fox_mie/agent-ready ci --comment
npx --yes @eshen_fox_mie/agent-ready ci --comment --write
npx --yes @eshen_fox_mie/agent-ready ci --write

Supported Detection

Current detectors cover common JavaScript/TypeScript, Python, Ruby, PHP, C#, Java, Kotlin, Rust, Go, and monorepo repositories:

  • package manager: npm, pnpm, yarn, bun, pip, bundler, composer, dotnet, maven, gradle, cargo, go
  • commands: install, dev, start, build, test, lint, format, backend install, local services
  • docs: README, architecture docs, ADR directories, existing agent docs
  • CI: GitHub Actions workflows
  • monorepos: npm/pnpm workspaces, Turborepo, Nx, Lerna, Rush
  • frameworks/tools: React, Vite, Next.js, Next.js App Router, Vue, Nuxt, Astro, Svelte, SvelteKit, Express, NestJS, Playwright, Storybook, FastAPI, Django, Flask, Pytest, Rails, Laravel, .NET, Spring Boot, Rust web frameworks, Gin, Docker, Docker Compose
  • framework signals: package dependencies, framework config files, and common route/component conventions such as app/, pages/, src/routes/, and .astro files

See detector coverage for the ecosystem-by-ecosystem table of signals and inferred commands.

Design Principles

  • Local-first: no API key, no telemetry, no network calls.
  • Canonical instructions: keep AGENTS.md as the source of truth.
  • Agent-agnostic: generate small shim files for Claude, Cursor, Gemini, and Copilot.
  • Verification-first: agents should know how to run the smallest relevant checks.
  • Compact context: instructions should be short enough to fit naturally into agent context.

Roadmap

  • More framework detectors and fixture coverage
  • Release automation and provenance hardening
  • MCP server for editor and agent integrations
  • Repository badge automation
  • Benchmarks on real open-source repositories

Contributing

Good first issues:

  • Add a detector for a framework you use.
  • Add a fixture for a repository shape that currently scores poorly.
  • Improve a lint rule with a clearer fix suggestion.
  • Add support for another agent instruction target.

Run checks before opening a PR:

npm run check
npm test
node bin/agent-ready.mjs score --fail-under 90

See CONTRIBUTING.md and LAUNCH.md for contributor and launch notes.