what jevguard is
a pre-sign wallet firewall. paste an unsigned transaction, a walletconnect request, or eip-712 typed data. the app decodes it, simulates asset changes without your private key, asks typeSafe jev focused questions about a facts json, then code emits allow / warn / block.
it is not a price bot. other crypto tools guess candles. this one refuses approve(uint256.max) to a random address.
jevguard never takes custody. it never asks for a seed. a block cannot send a transaction.
the four-step loop
- decode selectors, abi, and eip-712. extract spender, allowance, max-uint, nft operators.
- contract intel: is it a contract? proxy? verified name?
- simulate asset changes. never needs a private key.
- build a compact facts sheet, ask jev noul/score/choice questions, apply hard rules, show a forensic report.
what jev sees
classification, shortened + full addresses, allowlist flags, allowance, simulation deltas, optional user intent. not the hex blob. not your seed phrase. not “should we sign this?”
jev answers atomic questions like “is this unlimited approval?” and “does this look like a drain?”. code still owns math. a jev “allow” cannot override a hard block.
hard rules
- unlimited approval to unknown spender → block
- setapprovalforall(true) to unknown operator → block
- eip-712 chainid mismatch → block
- multi-asset drain pattern in simulation → block
- permit + max amount + unknown spender → block
- eoa + unexpected calldata → at least warn
- deploy with value → warn
simulation
the scanner previews native and token movement, new approvals, and reverts. if a paid simulator is not available, it still runs a local call and reports that honestly. swaps without a full trace still show native out when the value field is set.
$jevguard + robinhood chain
the native token lives on robinhood chain (chain id 4663, gas is eth). it is the dashboard access pass. anyone can open /dashboard and look around. using scanner, token intel, history, or fixtures requires a connected wallet. when the token gate is on, that wallet must also hold the amount set in site settings. when it is off, connecting is enough. the publisher wallet is exempt. $jevguard is not the sample for summary, lore, or launch clusters.
token intel
token intel lives on the dashboard next to the scanner. paste any project's 0x contract, website, or x profile. jevguard returns a public summary, published lore if it exists, and launch-wallet clustering. scans also attach the same snapshot for tokens named in the transaction.
clustering looks for same-transaction distributions, same-block coordinated receives, and many wallets sharing one sender, drawn as a simple hub-and-spoke graph. it does not prove same-ETH-funding without a dedicated indexer. if a contract cannot be resolved from a site or profile, lore can still come from that page — clustering waits for a 0x address. missing fields stay missing; lore is never invented.
wallet intercept
connect with privy (injected wallets, walletconnect, and the usual mobile wallets). paste a walletconnect json payload if you already have a session request. after allow, send through your wallet. after warn, hold three seconds. after block, no send.
the report
every scan produces a verdict, the hard-rule hits, jev’s question answers, simulation deltas, and a one-sentence summary. copy the markdown if you want a record. history stays in this browser only.
threat model
simulations can be wrong. malicious contracts can behave differently on-chain. novel encodings can be missed. phishing sites can still trick you into holding “sign anyway” after a warn. not a substitute for a hardware wallet. block never auto-sends.