ERC-8004: Trustless Agents

Hi everyone — How are you doing ?

Building on @azanux’s point about standardized function names and @davidecrapis.eth’s IReputationRegistry / IValidationRegistry sketch —

ERC-8004 gives agents an identity. But identity without verifiable trust data creates the same problem we have with tokenized assets today: everyone has an address, nobody has a quality signal.

An agent registered via ERC-8004 currently has no standard way to:

  • Prove its track record across chains (847 jobs, 92% success rate — factual, not subjective)

  • Be held accountable when it causes a loss (who pays? deployer, operator, evaluator?)

  • Signal risk in real time when the system it operates in enters a crisis regime

The getAuthFeedback() and getValidationResponse() interfaces @davidecrapis.eth sketched are a step in the right direction. But feedback is opinion-based. What the ecosystem needs is factual, signed, contestable attestations — measurements, not ratings.

I’ve just submitted a draft ERC — Trust Infrastructure for Autonomous Agents and Tokenized Assets — that defines five composable interfaces designed to sit above ERC-8004:

  • IAttestation — factual quality measurements (score 0-1000, with decay and contestability)

  • IDecisionTrail — on-chain audit trail for every autonomous decision (MiCA/EU AI Act compliant)

  • IAccountability — dispute resolution with a decentralized tribunal (3 evaluators, 48h verdict)

  • IRiskSignal — cross-chain regime detection (STABLE → CRISIS), bridge-agnostic

  • IRWAPassport — 7-layer risk passport for tokenized assets

The framing: ERC-8004 = identity. This ERC = trust. An agent gets an 8004 identity, then builds a verifiable track record via IAttestation, traces its decisions via IDecisionTrail, and is insurable via IAccountability.

To @azanux’s interoperability concern: all five interfaces have standardized Solidity function signatures. getAttestation(address) → (score, confidence, timestamp, attester) is the same call on any implementation, any chain.

Reference implementation is live on Base (19 contracts) and Gnosis (7 canonical contracts), 439 tests, 120+ assets scored. Would welcome feedback from the 8004 community on the composability model.

— Patrick (ALIA Quality Network)