Your AI coding agent is not the problem. Your repo is.

You bought Claude Code or Cursor. Your team is faster on greenfield work and slower on anything that touches the real codebase. The agent invents a service that already exists, writes tests in a style you abandoned two years ago, misses the one file that actually had to change, or fixes a bug by adding a fallback that hides it.

The usual conclusion is that the model is not good enough yet.

Usually the model is fine and your repository is unreadable to an agent. The rules that make your codebase work live in your reviewers' heads. A senior engineer catches a violation in seconds. An agent cannot see the rule at all, and your test suite goes green anyway.

That is a diagnosable condition, and it is cheap to fix once somebody has actually read your code for it.

What you get

An AI-Readiness Codebase Audit. $149. Delivered within 24 hours of repo access.

You give me read access to one repository. You get back:

A findings report of 8 to 15 pages, ranked by what each finding costs you. Every finding names files and line numbers in your code, and states the concrete failure: the prompt a developer on your team would plausibly write, and what the agent would do wrong with your repo as it stands today.
A real CLAUDE.md or AGENTS.md, written from your code. Architecture, domain model, conventions, testing patterns, and the rules that currently exist only in review comments.
Subdirectory context files for the three to five areas where one root file is not enough.
A convention inventory: the rules your codebase actually follows, extracted from the code rather than from what somebody believes is true.
A test-signal assessment: which behaviours are pinned by tests and which are not, which is to say where an agent can break you silently.

See the work before you buy it

I audited dubinc/dub, a 24,000-star commercial codebase, for free and published the whole thing.

It found two link-creation functions that take the same type and diverge in six behaviours, an invariant enforced 500 lines away from the code that depends on it, and a test suite that pins the one path where nothing was broken. All with file and line references.

Read the full sample audit

Judge the work, not the bio. That is what the sample is for.

Price


First three clients	$99
Standard	$149
Each additional repository, same engagement	$99
Re-audit after you have acted on the findings	$99

The founding rate is genuinely the first three, not a rolling countdown. When it is gone this page will say so.

If the report contains nothing you did not already know, you pay nothing. No forms, no argument. I can offer that because a report full of line references to your own code is very hard to fake.

Scope, stated up front

One repository per audit. Up to roughly 150,000 lines. Larger monorepos are quoted separately or audited by subsystem.
Read access only. I do not push, do not open merge requests, and do not touch your CI.
Public repo, or a read-only deploy key or guest access on a private one. If your code genuinely cannot leave your side, this does not fit and I will tell you so rather than sell it to you.
This is not a security audit, not a performance audit, and not a review of your business logic. Those are different products. Saying so is what protects the 24-hour turnaround.

How it works

Email me the repository and what your team keeps getting wrong with it.
I confirm it is in scope and send a PayPal link.
You pay, and you send read access.
The report and the context files land within 24 hours as a markdown bundle.

Is this AI-generated?

Yes, and that is the product. An AI agent audits your repository for how well AI agents can work inside it, and I review and send the result. If that sounds circular, it is: the auditor has exactly the same blind spots as the agent you are trying to make productive, which is precisely what makes the blind spots visible.

I would rather tell you that on the pricing page than have you work it out afterwards.

Who I am

Sabelo Simelane. Technology and innovation leader, Johannesburg. I run this audit myself and answer every email personally.

Get an audit: email sabelo@ligabazi.co.za, or pay $149 (or $99 for the founding rate) to paypal.me/sabside and reply with your repo. Asking a question first is usually the better order.