inspect-receipts
Python package for emitting local ZAP1 receipt requests from Inspect eval task-end logs.
agent work, bounded by receipts
Agents can act. The useful question is what another person can verify later.
00ZEVEN is the public edge of a receipt-first operating style: bounded action, local evidence, ZAP1-compatible receipt, verifier output. No market promise. No custody. No hidden dashboard trust.
The workflow is deliberately small.
The agent runtime can be complex. The public proof object should not be.
01 observe
An eval, action, skill, or operator step produces local evidence.
02 bound
The claim is narrowed until it has a clear verifier and a clear limit.
03 receipt
Sensitive material stays local. The receipt carries hashes and policy context.
04 verify
A reviewer checks the receipt without trusting the issuer server.
What is live now.
The public on-ramp is not a broad agent platform. It is one installable receipt hook for Inspect AI evals.
zcash-mcp
Agent-facing ZAP1 receipt and verifier tooling. It is not a wallet or signer.
VWC discipline
Locked harness, tiny editable surface, validity before score, receipt after proof.
Boundary.
No market execution claimThis page does not claim venue fills, live order routing, or executed market activity.
Not a custody surfaceNo keys, funds, wallets, balances, or private account state are handled here.
Not an endorsement claimInspect, Hermes, Gray Swan, XBOW, and other ecosystems are not claimed as adopters unless they actually use it.
Not ZK correctnessCurrent agent receipts are hash commitments unless a real proof statement and verifier exist.