Agent-Ready Repo Auditor

Find the public setup, instruction, verification, and safety-boundary gaps that make coding agents unreliable.

Agent-Ready Repo Auditor is a dependency-free Python CLI, library, and GitHub Action for repositories used with Codex, Claude Code, GitHub Copilot coding agent, Cursor, or mixed-agent workflows. It reads one immutable public default-branch snapshot and emits deterministic JSON or Markdown with source links for every positive finding.

Request a free automated audit · Review the $750 human-reviewed audit · Review the $249 Instructions PR · Review the $149 Fix Plan · Review the $495 AI Agent Cost & Reliability Snapshot · Inspect the real two-file sample PR · See the immutable audit report · Install the GitHub Action

The four paid-offer pages are public and require no login. GitHub sign-in is required only when opening a public request.

It never clones or executes repository code. It is not a vulnerability scanner, security assessment, compliance review, or proof that documentation is accurate.

GitHub Action

Add a deterministic public-evidence check to a repository:

name: Agent readiness

on:
  pull_request:
  push:
    branches: [main]

permissions:
  contents: read

jobs:
  audit:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: wrightops-ai/agent-ready-repo-auditor@v1
        with:
          fail-threshold: "65"

The action writes agent-ready-audit.md, adds the evidence report to the workflow summary, and exposes score, band, report-path, and immutable revision outputs. Its token is used only for bounded read requests to api.github.com; private repositories still fail closed.

Want to see the output before installing it? Request one free automated audit. Requests and reports are public. The form's remediation-interest question is non-binding demand research; opening an issue does not create a service contract.

Need exact next steps after the report? The $149 Agent-Ready Repo Fix Plan adds exactly three human-reviewed fix cards to the public audit issue within one business day. Submit the completed audit for automated preflight and scope confirmation, then use the dedicated PayPal goods/services checkout only after WrightOps confirms that the public request fits the fixed scope.

Need a human-reviewed evidence audit and handoff? The $750 Agent-Ready Repository Audit covers one public repository at one immutable revision, evidence-linked JSON and Markdown, up to five priorities, and a 30-minute handoff. Request a written scope decision without posting contact or payment information. Opening the issue is not a contract or payment obligation; WrightOps confirms scope before privately sharing a PayPal Goods & Services checkout, and work starts only after provider-confirmed settlement.

Choosing which instruction file should own which rules? Read the no-login AGENTS.md vs CLAUDE.md vs Copilot Instructions field guide. It is educational and requires no login; reading it is not a customer request, contract, or payment obligation, and the guide is not a security or compliance assessment.

Need the two instruction files proposed as an actual contribution? The $249 Founding Agent-Ready Instructions PR adds or updates exactly root AGENTS.md plus one selected companion file—root CLAUDE.md or .github/copilot-instructions.md—for one public repository, then delivers one ready-to-merge pull request with one revision round. It uses only committed public evidence and does not execute repository code. Request written scope confirmation; the dedicated PayPal Goods & Services checkout is shared privately only after scope confirmation. An eligible $149 Fix Plan for the same repository and revision can be credited toward the $249 total. Inspect the merged sample PR to see the Copilot companion variant, exact two-file boundary, evidence notes, hosted checks, and public review surface.

Need aggregate cost, reliability, latency, and retry evidence instead? Review the $495 AI Agent Cost & Reliability Snapshot and its public terms, then use the structured public scope request only after confirming the prompt-free input boundary. The request is non-binding; WrightOps confirms scope before sharing any checkout.

For a concrete example, read the sample report for the immutable v1 release, generated from public evidence at revision 7a507bc0cb42f8c04fb18e53a46371b37b5bd56f.

Setup

Prerequisite: Python 3.10 or newer. There are no third-party runtime dependencies.

No runtime environment variables are required. The GitHub Action accepts an optional github-token input, used only to raise API rate limits for public read requests.

cd agent-ready-repo-auditor
python3 -m agent_ready_repo_auditor owner/repository --format markdown

Write both common delivery formats:

python3 -m agent_ready_repo_auditor owner/repository --format json --output audit.json
python3 -m agent_ready_repo_auditor owner/repository --format markdown --output audit.md

Accepted inputs are exactly owner/repository or a canonical https://github.com/owner/repository URL. Private, missing, inaccessible, or GitHub-truncated repositories fail closed without a partial score.

Evidence model

The 100-point evidence score covers:

Public evidence	Points
Root README with setup guidance	20
Recognized coding-agent instructions	20
Redacted environment example or explicit README no-configuration statement	10
GitHub Actions workflow	15
Issue and pull-request templates	10
Verification commands or automation	15
Explicit risky-action boundaries	10

Missing evidence means only that the artifact was not evidenced in the public default-branch snapshot. The report includes the inspected revision SHA, evidence URLs, limitations, warnings, and prioritized next steps.

Recognized agent instruction files are root AGENTS.md, root CLAUDE.md, .github/copilot-instructions.md, and root .cursorrules. Risk-boundary credit requires explicit prohibition or approval language tied to risky actions; a generic mention is not enough.

Library usage

from agent_ready_repo_auditor import audit_repository, render_json

report = audit_repository("owner/repository")
print(render_json(report), end="")

Successful JSON uses schema agent-ready-repo-audit.v1. Expected failures have a stable error object on standard error:

{
  "error": {
    "code": "repository_inaccessible",
    "message": "Repository is private, missing, or otherwise inaccessible through the public GitHub API.",
    "retryable": false
  },
  "ok": false
}

Unauthenticated GitHub API limits apply to the CLI. The GitHub Action uses the workflow token for bounded API reads. Rate-limit failures include reset evidence when GitHub supplies it and never emit a partial audit.

Verification

All automated tests use a mocked HTTP transport and make no network calls:

python3 -m unittest discover -v
python3 -m py_compile agent_ready_repo_auditor/*.py tests/*.py

An optional live smoke test can be run against any public repository:

python3 -m agent_ready_repo_auditor wrightops-ai/website --format json

Commercial delivery boundary

This repository is suitable as the deterministic evidence engine for a human-reviewed readiness audit. A responsible paid delivery should state the scanned revision, manually verify the generated findings, add repository-specific priorities, and preserve the report's limitations. Do not market it as a security audit or promise operational outcomes that were not measured.

The primary human-reviewed offer is the $750 Agent-Ready Repository Audit. Its structured public scope request records requester authority, one public repository, the workflow or pain to prioritize, the handoff preference, and explicit scope acknowledgements without collecting contact or payment details. Opening the issue does not create a contract or payment obligation. WrightOps confirms the immutable revision, scope, acceptance criteria, timing, and exclusions before privately sharing a PayPal Goods & Services checkout; human review starts only after provider-confirmed settlement.

The fixed-price entry offer is the $149 Agent-Ready Repo Fix Plan. It covers one public repository and exactly three human-reviewed fix cards, delivered publicly on the completed audit issue. It does not include implementation. A structured public request records authorization and scope acknowledgement without collecting payment details. Automated preflight verifies that the audit and repository match; use the dedicated PayPal goods/services checkout only after WrightOps confirms the fixed scope.

The implementation offer is the $249 Founding Agent-Ready Instructions PR. It covers one completed audit or no-login repository preflight, one public repository and immutable revision, exactly root AGENTS.md plus one selected companion file—root CLAUDE.md or .github/copilot-instructions.md—one ready-to-merge pull request, and one revision round. WrightOps manually revalidates the public repository and immutable revision before confirming scope, uses only committed public documentation and workflows, and does not execute target repository code. Its public request records the public inputs and authorization; written scope confirmation comes before the dedicated PayPal Goods & Services checkout is shared privately. The merged sample PR shows the Copilot companion variant of the promised two-file contribution on an owned public repository.

License

MIT