Blog
AI Software Factory
From the article

Your app works. But what's actually in there?

The audit prompt from the article. Paste it into Claude Code, Cursor, or Codex. Find out what you don't know is in there.

Get the free audit prompt

No spam. Unsubscribe any time.

audit-findings.md Read-only 
## 1) Executive summary

Not safe for autonomous AI coding.

Highest risk: two competing auth paths with no canonical
authority declared. An AI agent will pick one confidently
and be wrong.

Problem type: documentation drift + unclear authority.

## 8) Security exposure findings

HIGH  api/webhooks/stripe.ts
      Signature verification missing. Any POST executes
      without validation. Exploitability: High.

MED   lib/supabase/admin.ts
      Service-role key imported in three non-admin paths.
      Blast radius if misused: Critical.

## 17) AI Change Governance Need Score

  Canonical path clarity          3/5  meaningful drift
  Auth/security boundary clarity  4/5  high risk
  Test/verification credibility   3/5  meaningful drift
  AI-worker safety                4/5  high risk

Overall: High

Illustrative output — your repo will produce its own findings.

Get the prompt ↑

What it uncovers

  • Which parts of your repo are unclear, duplicated, stale, or risky
  • Where an AI agent is most likely to extend the wrong path
  • Whether auth, ownership, and data boundaries are obvious enough to trust
  • Whether your tests and CI actually prove safety
  • Whether your repo needs stronger governance before more AI coding

What you receive

AI Architecture & Security Drift Audit

Open your repo in a repo-aware AI coding tool, paste the full prompt, and get a structured findings report — including an AI Change Governance Need score.

  • No code changes
  • No installs
  • No migrations
  • No fixes
  • Read-only findings report

This is not a security guarantee or a full code review. It is a static, read-only diagnostic designed to reveal whether your repo has enough architecture clarity, security boundaries, and workflow discipline for safe AI-assisted development.