Every AI agent is a silo. The Passport is shared context that lives in git and travels across agents/vendors — and a poisoned/injected entry is screened out before any agent trusts it. Try the trust screen: paste a context entry an agent might leave.
In your repo: npm i -g mneme-ai then mneme ctx inherit / contribute · MCP mneme.context.inherit (every agent) · Suite · TRUST-precision 1.0 — poison never inherited · honest: screens known poison + grounds on citations, not a proof of truth.