Hi everyone,
I’ve been experimenting with a minimal governance pipeline for AI agents.
The idea is simple: if agents are going to execute real tasks and call tools, we need a way to audit what actually happened.
This demo demonstrates a minimal architecture:
Persona (POP)
→ Agent Runtime
→ Execution Trace
→ Audit Evidence (ARO-style record)
Key points:
• deterministic demo (no API key required)
• CrewAI agent workflow
• append-only audit evidence
• minimal governance pipeline
Repo name:
joy7758 / verifiable-agent-demo
Run locally:
bash scripts/run_demo.sh