Trust tiers

The scam problem on an open chain is a naming problem. Anyone can deploy a token called USDC or TSLA. Robinhood Chain's explorer lists counterfeit versions of major stocks and a fake Global Dollar — same names, worthless contracts.

pouch therefore never trusts a name. Before the agent can touch any token, the address is classified into a tier using signals that cost money to fake.

The tiers

Official

Reserved for the ~95 official Robinhood stock tokens (AAPL, NVDA, TSLA, and the rest).

Checks, all on-chain unless noted:

• Bytecode fingerprint. Every official stock token shares the exact same runtime bytecode hash. The candidate's code hash must match it.
• Deployer cross-check. When the explorer API responds, the contract creator must be the official Robinhood issuer address. A matching-bytecode token from a different deployer is flagged as a suspected spoof and demoted to unknown.
• Live liquidity. At least one live Uniswap v4 pool against a quote currency.

Issuer

Tokens deployed by an issuer you explicitly trust, via the policy.trustedIssuers registry. Useful as new RWA issuers come onchain: add their verified deployer address, and their tokens classify above established with their own per-trade cap ($250 by default).

Empty by default — nothing classifies here until you opt an issuer in.

Established

Real tokens with real markets.

• 1,000+ holders (threshold configurable via minHolders)
• An indexed price feed (CoinGecko-linked exchange rate on the explorer)
• Live Uniswap v4 liquidity against ETH or USDG

Faking a name is free. Faking thousands of holders, a listed price feed, and deep pool liquidity is expensive.

Unknown

Everything else. Blocked by default. This is where the counterfeit stablecoins and fake stock tokens end up.

A real example

Search TSLA and several results come back. Only one passes the official checks: bytecode fingerprint match, verified deployer, live USDG pool. A copycat with the same name and even a plausible holder count fails the fingerprint check and classifies as unknown — blocked before a quote is ever fetched.

Honest limits

Tiers make scams expensive, not impossible. Copied bytecode plus genuinely seeded liquidity could fool the official tier while the explorer's creator-check is unavailable. Treat tiers as a strong filter, not a guarantee, and keep the pouch small. See Security model.