Agent execution protocol
The machine-checkable rules under which an AI agent is permitted to do consequential work: authorization, jurisdiction, escalation, and how finished work seals into a verifiable record.
An agent execution protocol answers, in a form software can check, the four questions any institution asks of anyone holding credentials: who authorized this agent, what may it act on, which decisions must escalate to a human, and where is the record of what it did. Without one, every answer is a convention in someone's head — and conventions do not survive a fleet.
The AAO Stack publishes its protocol layer as open source: the AAO manifest format and conformance suite, the agent SDK, and the settlement verifier, all Apache-2.0 and indexed at flashyos.com/open. Completed cross-organizational work seals as canonical JSON plus its sha256 — a record a stranger can recompute rather than one the operator asserts.
The operator definition lives at flashyos.com.
Common questions
How is this different from an agent communication protocol like A2A or MCP?
Communication protocols move context and requests between agents. An execution protocol governs whether the work itself is permitted — authorization, approval thresholds, and settlement. You need both; only one of them answers to an auditor.
Can I adopt the protocol without the network?
Yes — that is the point of shipping it open. The manifest format and conformance suite run in your own CI under Apache-2.0. The network that enforces it live is the product.