AI Delivery Risk Audit - 7-Day Report on Your Agent Codebase
$2500.00
A 7-day, human-verified audit of one repository against the seven structural failures that agent-built codebases share. You get a scored report, a priority stack, and a recorded walkthrough of your ow
Buy on Gumroad
← All products
What this is
A 7-day, human-verified audit of one repository against the seven structural failures that agent-built codebases share. You get a scored report, a priority stack, and a recorded walkthrough of your own code.
This is not a tool download. I scan your repository, then verify every finding by hand and delete anything that does not hold up. A report that never excludes anything is not an audit.
WHAT YOU RECEIVE, 7 DAYS AFTER REPOSITORY ACCESS
- Scored PDF report: 7 checks, 0-100 score, one evidence line per judgment (file paths and counts you can re-run)
- Priority stack: what to fix first, ranked by score gap x weight, with hours attached
- 45-minute recorded walkthrough, screen-shared, finding by finding in your code
- One 30-minute live Q&A call within 14 days
- One free re-scan after your fixes, to confirm the score actually moved
THE SEVEN CHECKS
1. Scope Containment - what stops the agent from editing files outside its mandate
2. False-Green Detection - can your tests tell "it ran" apart from "it is correct"
3. Drift Containment - is "done" written down where the agent can read it
4. Memory Integrity - when two sessions disagree, who arbitrates
5. Delivery Gate - what blocks 80% done from being reported as done
6. Complexity Control - are files so large that the agent's context truncates mid-edit
7. Regression Surface - after one change, can you enumerate what else moved
EVIDENCE THIS METHOD FINDS REAL THINGS
Run against three mature public repositories with the same method:
pydantic/pydantic-ai 80.2/100 - openai/openai-agents-python 77.8/100 - microsoft/autogen 55.2/100
Two of those carry 97 and 50 business-logic files over 800 lines. Well-funded teams, clean CI, and the exposure is still there.
WHAT YOU GET AT CHECKOUT
A start kit containing the intake form, the scope and terms, the report template, and three complete sample audits on real public repositories - so you can judge the output before you send me anything.
WHAT I NEED FROM YOU
One repository URL you own or have written permission to analyse. Private repos: I send a read-only deploy key. The delivery clock starts the day the repository is accessible, not the day of purchase.
SCOPE
One repository, up to 500k lines. Static structural analysis only - it does not read your application logic for correctness bugs, does not run your code, and is not a security penetration test.
GUARANTEE
If the report does not contain at least three findings you consider actionable, say so within 7 days of delivery and you get a full refund. No forms, no qualification questions.
What ships
How you get it
One-time payment of $2500.00. Instant download from Gumroad after checkout — no account on this site is needed. VAT/GST/sales tax is handled by Gumroad as merchant of record. MIT licensed for commercial use and modification with attribution.