Trust Center
Trust is not a label — it is a system of verifiable artifacts. This page documents the safeguards active in the protocol today and the boundaries that still exist while the trust layer matures toward on-chain enforcement.
ERC-721 token on Base Sepolia links each agent to a verifiable on-chain identity
Operator ownership enforced — users only see and control their own agents
Strategy type, risk profile, and execution wallet recorded as immutable metadata
Dynamic position sizing: 10% / 25% / 40% of treasury based on risk profile
Daily loss limits, kill-switch drawdown thresholds, and duplicate exposure blocking
Low-conviction entries (score < 65) auto-cut at -1.5% instead of waiting for stop-loss
Every trade intent scored 0–100 with risk checks, AI reasoning, and checkpoint data
Validation timeline grouped by nonce — searchable, paginated, and timestamped
Blocked trades produce explanation artifacts showing which policy was violated
Firestore security rules enforce field-level validation and document ownership
Deep undefined stripping on all writes prevents silent data corruption
Unique nonces per cycle with random suffix prevent grouping collisions
Forge8004 enforces trust through structured intent validation, pre-trade risk routing, checkpoint persistence, and owner-scoped data access. Every agent action is auditable from decision through outcome.
The protocol is evolving toward fully on-chain enforcement — capital vault contracts, trustless risk routers, and DEX-settled execution on Base Mainnet. Today it operates as a verifiable sandbox with real market data and AI-driven decision logic.
Trade execution is currently simulated — not routed through a live on-chain settlement contract.
EIP-712 typed intents and EIP-1271 smart-wallet verification are planned but not yet implemented.
The risk router enforces rules in application logic today. On-chain enforcement is the Phase 4 goal.
API keys and private keys are server-side only — never exposed to the client bundle.