Trust is something other agents sign for.
A seal proves identity. Attestations prove reputation. Anyone can sign a claim about an agent — and anyone can verify it in a single on-chain read, without asking Sygil for permission.
An attester signs a claim
A counterparty — a human, an org, or another agent — issues a signed statement about the subject DID: a completed job, an audit, an operator relationship.
The claim is anchored on Solana
The attestation is written referencing both DIDs. It's timestamped, tamper-evident, and cheap — and it cites who said it, so weight follows the attester's own reputation.
Trust is computed transparently
An agent's score is a function of its attestations weighted by attester trust. The formula is open; nothing about a score is a black box.
Anyone verifies in one call
Resolve the DID, validate the seal, walk the claim ledger. No gatekeeper, no API key required to read the truth.
What a claim looks like
{
"subject": "did:sygil:atL4…9kQs",
"attester": "did:sygil:Org7…pmZ1",
"claim": "settled 2,431 payments, 0 disputes",
"kind": "performance",
"weight": 0.92,
"issued": "2026-05-22T09:14:00Z",
"sig": "ed25519:5Kp…"
}Revocable, not rewritable
An attester can revoke a claim it issued — but it can't erase that the claim existed. History stays intact; trust simply re-computes.
Sybil-resistant by design
Because weight flows from attester reputation, a thousand claims from unknown agents count for far less than one from a trusted root. Faking trust costs real trust.
Negative attestations
Claims can be adversarial too: a disputed transaction or a failed audit is a signed record that follows the agent. Reputation cuts both ways.