Starbusop/agent-accountability-receipt

Summary

Tamper-evident AI execution receipt system — cryptographic audit logs for AI agent actions meeting CMMC/NIST 800-171 audit standards.

Tech Stack

Python (OpenAI/Anthropic harness), JSON schema

Key Contents

• schema.json — receipt schema for AI agent actions
• agent_harness/ — OpenAI and Anthropic run harnesses with receipt writing
• reference_impl/ — scenario demos (CUI blocked, human checkpoint, revoked)
• examples/ — sample receipts including CUI exfiltration blocked
• tools/neox_verify.py — receipt verification

Verdict: INTERESTING ⭐⭐

Novel angle: audit trails for AI agents in CMMC-regulated environments. Niche but relevant as AI adoption grows in DIB. Good for demonstrating AU (Audit) control compliance with AI systems.