Status: Independent Technical Working Draft
Version: 0.9.1
Review Status: External technical review requested
Agent-to-Secure Payload Authorization (A2SPA) defines an execution-time authorization model for determining whether a specific structured execution request may be admitted immediately before execution.
An Execution Request SHALL NOT be admitted to the Execution Environment unless the Execution Enforcement Point has received a successful Authorization Decision from the Authorization Verifier for the corresponding Signed Canonical Execution Request.
A2SPA addresses deterministic, cryptographic validation of structured execution requests at the execution boundary.
It does not validate natural-language prompts, model alignment, semantic correctness, business correctness, legality, or the desirability of an upstream decision.
Feedback is welcome on:
- Security assumptions and threat-model gaps
- Canonicalization and signature interoperability
- Replay, nonce, TTL, and failure-state handling
- Deployment and bypass risks
- Conformance requirements
- Implementation concerns and edge cases
- Architectural flaws or falsification conditions
Please open a GitHub Issue with the relevant section number and severity: blocker, high, medium, or low.
This repository is an independent technical working draft. It is not an IETF working-group document, adopted standard, certification, or security audit.