onchain_agent_preflight

ai.satohub/onchain-agents · Verify this server

USE WHEN you are about to install a package, clone a repo, connect to an MCP endpoint, pay an agent, or trade a token, and you want to know what is on record about it FIRST — every evidence line, with the field it was read from and its date. Pass exactly one of repo, package, endpoint, agent, token, skill, x402 or address; with none (or a template placeholder) nothing is checked and the answer is { needs_input: true, accepted_forms, example_calls }. NOT FOR a stack pick's summary (onchain_agent_recommend_stack and onchain_agent_get_deploy_spec already carry a records-only `preflight_summary`), whether an install command or MCP config takes your key (use onchain_agent_check_install instead), or a recipient-address reading on its own (use onchain_agent_scan_recipient). Call this for the full evidence, for any target that is not a pick (a repo, package, token, skill or ERC-8004 agent you were handed), or for a live handshake with an unlisted MCP endpoint. RETURNS a verdict — go | caution | no | unknown — with one evidence line per check, each naming the field it was read from and when that field was written. The rules are written down in lib/preflight.ts and cited by id in `rule`. RULE ENFORCED: a verdict names WHAT WAS CHECKED AND WHEN. It is never a security review, a quality judgment or a statement about returns, and `unknown` means Sato Hub holds no record — not that something is wrong. An unlisted endpoint gets ONE live handshake (initialize + tools/list, 8 s cap) and can never come back `go`: a handshake is not a record. For agent=<chain>:<id> we confirm the ERC-8004 registration exists, fetch its registration file, and report the services it DECLARES; only a declared MCP service is probed. TOKEN LANE (token + chain; the EVM chains listed, and Solana): keyless chain reads. EVM: bytecode presence and size, the ERC-20 views, the Clanker v4 factory's OWN deployment record, and the Uniswap v3 factory across the four standard fee tiers against wrapped native. Solana (token = the mint): token program (SPL or Token-2022), decimals, supply, mint and freeze authority (revoked is a known fact, distinct from unread), Token-2022 extensions (PermanentDelegate, TransferHook, transfer fee, NonTransferable), the ten largest token accounts as a share of supply, and pump.fun bonding-curve state. A PermanentDelegate, TransferHook or transfer fee is a caution; a NonTransferable mint is a no; there is no `go` for Solana. Every field is nullable and a null carries its reason. Holder concentration on EVM and Uniswap v4 pools are permanently null and said so. A pool existing is not depth. Nothing in this lane says safe, audited, rug or scam: those are not readings. SKILL LANE (skill): a skill is a DOCUMENT an agent follows, which is why it is worth checking first. Evidence is the static disclosure the weekly sweep produced: hosts the text names, whether it generates or handles keys, asks for a credential, pipes a remote script into a shell or grants itself tools, each finding WITH the lines that produced it. Nothing is fetched from a registry and no skill is executed. A DISCLOSURE DESCRIBES: it never says safe or malicious, and an empty flag list is "nothing matched", not a pass. RETURNS (json): { verdict, rule, target: { kind, value, slug, name, sato_url, verify_url }, evidence: [{ check, result, source_field, checked_at }], checked_at, caveat, rules, token?, skill?, solana_receipt? }. `token` and `skill` are the raw reports for those lanes; `solana_receipt` (present with a custody reading) is where that reading is recorded on Solana, or null. Read-only. X402 LANE (x402): one unpaid request to the resource URL; reports whether it answered 402 and the payment terms it states (network, asset, amount, payTo), against its catalogue entry and our daily record when we hold one. It describes the payee — it is never a reputation score. CUSTODY: repo, package, endpoint and skill results carry `custody` (Sato Check: does it take your key, does the key leave, can it move funds, what changed) when a profile is on record; custody can only lower a verdict, never raise it. ADDRESS LANE (address + chain, optional origin and from): the Sato Scan recipient check before a payment, five dated questions (look-alike, issuer's token contract, declared payee, structural mark, address poisoning), rules A1-A7. `go` is declared, canonical and nothing found within the stated limits, never a clearance. Nothing is stored. The reading on its own: onchain_agent_scan_recipient. Example: { repo: "coinbase/agentkit" } · { endpoint: "https://mcp.example.com/v1" } · { x402: "https://api.example.com/paid" } · { agent: "base:42" } · { token: "0x1bc0c42215582d5A085795f4baDbaC3ff36d1Bcb", chain: "Base" } · { skill: "clawhub/solana-wallet" } · { address: "0x1bc0c42215582d5A085795f4baDbaC3ff36d1Bcb", chain: "Base" }