Vendor-neutral attestation format for agent reputation. Ed25519-signed, 100% verification fidelity.
The problem
A2A defines how agents talk. It doesn't define how they come to trust each other. Today every framework has its own ad-hoc reputation model, or none. Agents can claim anything about themselves; verifiers have no portable way to check claims from other parties.
The proposal
agent-reputation is a vendor-neutral attestation format: any agent can sign a claim about another agent's competence, and anyone can verify the claim using the public key embedded in the attestation. No shared secrets, no registry needed. Ed25519 signatures, deep-canonical JSON, 100% verification fidelity verified on 30 hand-built samples, three byte-identical runs.
Why now
A2A and AGNTCY are crystallizing. Multi-agent frameworks are shipping. Trust is the missing primitive. The right moment to plant a vendor-neutral attestation format is now — before each framework ships its own incompatible version.
Get notified when agent-reputation hits 1.0.
Who's behind this
Built by Aiden — solo developer working through a portfolio of AI primitives in 2026. agent-reputation is one of several exploratory bets; see the hub for the others. This is alpha — expect rough edges, please file issues.