The TrustOps Framework
The operational discipline for running AI trust continuously — the way modern teams run security, cloud and software. Trust established once will drift; TrustOps keeps it true.
The Metamynd Trust Index
A signed, portable trust score per agent, built on the HCS-28 attribution shape (weighted adapters + deterministic-denominator weighted mean). It's computed from what Metamynd uniquely holds — the verified owner, every allow / observe / block / escalate decision, spend/settlement behaviour and anchored evidence — and is guidance, not a silent gate: a low counterparty score routes to human review, never an auto-block.
Continuous operations
- Trust monitoring — the live index + coverage
- EvidenceOps — Merkle-batched, HCS-anchored, offline-verifiable
- GovernanceOps — edit Standards & SOPs live, no redeploy
- Escalation review — human-in-the-loop approval queues
Measurement
- Trust score & coverage per agent
- Decision compliance rate
- Escalation discipline
- Mandate adherence & settlement integrity
- Evidence anchoring
Issue and consume trust
TrustOps runs the loop both ways: Metamynd publishes each agent's trust record to an HCS-28 transparency topic (anyone can re-derive and verify it offline), and consumes the ecosystem's trust — a counterparty's HCS-28 score — to guide escalation. Governance data feeds the ecosystem's trust, and the ecosystem's trust guides review.