Auction Risk and Compliance Scoring Agent
Code available as a research and human-review screening tool, not a production misconduct detector.
An evidence-first Python screening agent that looks for unusual auction patterns and creates explainable reports for human review without declaring fraud.
- Why
- I wanted to turn a concerning auction experience into a systematic way to distinguish unusual patterns from misconduct.
How I built it
- I wrote rules separating observed bids, computed indicators, alternative explanations and compliance questions.
- I specified persistence, coordination-like patterns, timing, association and rules-related signals.
- I used structured schemas, transparent weights and tests for the deterministic scorer.
- I prohibited scraping, automated bidding and unsupported accusations.
- To copy it you need
- Start with synthetic bid records and a plain-language explanation of what the scoring system may and may not conclude.
Try it yourself
Copy this prompt and paste it into ChatGPT or Claude to try it yourself.
I want to build a small version of Jovan Diaz's Auction Risk and Compliance Scoring Agent from his What Are We Building? profile. It reads sequences of bids and flags unusual patterns for a person to review. It never says anyone cheated. His method: keep what was observed, what was calculated, the other possible explanations and the open questions apart, and keep the scoring weights in plain view. His gotcha: a rotation that looks suspicious can be a legitimate automated or round-robin bidder. Use invented bidders only. Start by asking which patterns I want to flag, then help me draft the columns and a few made-up test cases. Ask me for anything you need.
Tools I used for this project
ChatGPT: I used ChatGPT to draft, test, critique or work through the workflow.
Codex: I used Codex to help build or revise repository files and tests.
Python: I used Python scripts or validation tools to process and check structured outputs.
GitHub: GitHub holds all of my repositories: code, instructions, changes and review history.
