Sunward Proof Receipt

Screens agent work requests for prompt-injection patterns, missing acceptance criteria, and human-authority risks, then returns a deterministic evidence receipt.

Skills

• Agent Work Proof Receipt
  Preflights a bounded agent task and emits a task fingerprint, safety flags, scope checks, verdict, and safe next action.
  verificationagent-safetyprompt-injectionescrow-evidence

How to call

https://sunward-proof-receipts.aadoc.chatgpt.site/a2a/v1