ArkForge Proof Specification — open standard for verifiable agent-to-agent execution proofs
proof-spec is an early-stage Python project in the AI payments / x402 ecosystem, focused on agent-to-agent, cryptography, ed25519, open-standard. It currently has 1 GitHub stars and 1 forks, and sits alongside related tools like trust-layer, peac, a2a-secure, SDAP, blocklace-a2a, voidly-pay.
Open standard for verifiable agent-to-agent execution proofs.
A deterministic proof format that binds a request, response, payment, and timestamp into a single SHA-256 chain hash. Anyone can verify a proof without ArkForge's code or infrastructure.
Given a proof JSON, verify it in one line:
printf '%s' "${REQUEST_HASH}${RESPONSE_HASH}${PAYMENT_ID}${TIMESTAMP}${BUYER}${SELLER}${UPSTREAM}${RECEIPT_HASH}" \
| sha256sum | cut -d' ' -f1
UPSTREAM and RECEIPT_HASH are empty strings when absent from the proof.
If the result matches proof.hashes.chain, the proof is intact.
| Implementation | Language | Status |
|---|---|---|
| ArkForge Trust Layer | Python | Reference implementation |
Want to add yours? Open a PR.
test-vectors.json contains 9 test cases (7 legacy string-concatenation vectors + 2 canonical-JSON vectors for spec_version 1.2/2.1). Any conformant implementation MUST pass all vectors.
The proof-spec is the cryptographic foundation for the Compliance Receipts v0.1 spec drafted by the qntm Working Group.
Field mapping:
| proof-spec field | Compliance Receipts v0.1 field | Notes |
|---|---|---|
proof_id |
receipt_id |
Same role: unique receipt identifier |
hashes.request |
input_hash |
SHA-256 of canonical JSON (identical algorithm) |
hashes.response |
output_hash |
SHA-256 of canonical JSON (identical algorithm) |
timestamp |
step.timestamp |
ISO 8601 UTC |
parties.agent_identity + parties.agent_identity_verified + DID endpoint |
step.agent_did |
agent_identity carries the verified DID when agent_identity_verified: true; did:web:trust.arkforge.tech resolves to the Ed25519 signing key |
arkforge_signature |
signature |
Ed25519 — signs the chain hash (proof-spec) vs. canonical receipt (CR v0.1) |
hashes.chain |
no equivalent | Additional witness: binds payment + identity + request + response into one sealed hash |
Fields in Compliance Receipts v0.1 with no proof-spec equivalent (pipeline_id, step.index, step.role, previous_receipt_hash, policy) are pipeline context provided by the calling agent, not derived by the proxy.
The proof format will evolve to support third-party provider attestations and multi-PSP payment verification. See the Trust Layer roadmap for the full architecture.
Certifying proxy for agent-to-agent transactions — SHA-256 proof chain, Ed25519 signatures, RFC 3161 timestamps, Sigstore Rekor
Portable signed records for automated interactions.
End-to-end encrypted communication for AI agents. The security layer that Google A2A forgot.
Secure Digital Agent Protocol — HTTPS for AI agent communication. Cryptographic identity, mutual authentication, encrypted payloads, trust delegation, and audit trails for multi-provider agent ecosystems. Works with A2A, MCP, and any agent-to-agent protocol.
Cryptographic audit layer for A2A: hash-chained, signed blocks with equivocation detection.
Off-chain credit ledger + hire marketplace for AI agents. Ed25519-signed envelopes, atomic settlement, hire-and-release escrow. https://voidly.ai/pay
A local-first AI agent with persistent memory, emotional intelligence, and a peer-to-peer skills economy.
The trust layer for agent-to-agent commerce — natural-language mandates, ERC-7710 delegated permissions, x402 payments, escrow, and dispute resolution as one open, catch-all Agent Skill / Claude Code plugin.
Agent behavior that compiles
The A2A x402 Extension brings cryptocurrency payments to the Agent-to-Agent (A2A) protocol, enabling agents to monetize their services through on-chain payments. This extension revives the spirit of HTTP 402 "Payment Required" for the decentralized agent ecosystem.
Aser is a lightweight, self-assembling AI Agent frame.
Golang SDK for A2A Protocol