Security architect. 23 years, ex-RSA Security
I build the deterministic layer that sits between an AI agent and the things it can actually do.
*[ACP](https://github.com/yacine-kellib/agent-control-plane)* — authorization for AI agents. The model proposes, it never commands. Formally verified, open spec, every claim replays locally.
*[tack](https://github.com/tack-run/specs)* — prepaid billing and control for AI products. Spend is metered and debited before each turn, not reconciled after.