I built a small deterministic policy engine to govern actions proposed by LLMs before they are executed.
The system enforces:
- fail-closed execution - signed out-of-band approval proofs - anti-replay protection - verifiable audit chain
The repository includes a CLI demo showing how irreversible actions are denied, approved, and audited.
Repo: https://github.com/nayfly/aetherya-core
Feedback welcome.