AnteSig is a preflight evidence and explanation layer. It is not a wallet, signer, transaction executor, safety oracle, or authorization system. The application may determine that evidence requires stopping or that a result may proceed to manual review. It cannot authorize a transaction.
Private keys, seed phrases, signing, and Monad mainnet transaction submission are outside the project boundary. Future code must not collect, persist, log, transmit, or request signing material.
The project also does not guarantee execution success, economic outcome, protocol correctness, or transaction safety.
- A token address is the asset identity. A symbol, name, icon, or decimal label is display metadata and cannot establish identity.
- Protocol, token, and spender addresses require Maintainer confirmation before use. The source and network context must be reviewable.
- An address inferred from prose, a screenshot, token symbol, model response, or unverified example is not approved.
- An approval spender must match the Maintainer-confirmed protocol expectation
and the user intent. An unexpected approval requires
STOP.
The decision must be STOP when any of the following occurs:
- any Warning is present;
- any transaction or simulation step rolls back;
- Receipt generation, parsing, or validation fails;
- complete Change coverage cannot be proven;
- Change or Receipt ordering cannot be proven;
- the simulated state chain is interrupted or a dependent step lacks the prior state it requires;
- observed execution is inconsistent with user intent, including account, asset address, direction, amount, slippage, protocol, spender, or recipient;
- an unexpected Capability, approval, call, or funds movement appears;
- the original Capability Tree was modified or its integrity cannot be shown;
- protocol, token, or spender identity has not been confirmed by the Maintainer; or
- any critical evidence is missing, ambiguous, stale beyond its stated context, or cannot be attributed to its source.
A system error or inability to complete a mandatory check is missing evidence
and therefore results in STOP.
Natural-language explanations, model output, UI labels, and summaries are not on-chain evidence. They may describe evidence only when the underlying source remains accessible and the description does not add certainty or meaning.
Raw Capability, simulation, Receipt, Outcome, Warning, gas, coverage, ordering, and state-continuity evidence must be preserved separately from display models. Presentation code must not repair, suppress, reorder, or reinterpret evidence to avoid a stop condition.
Quote data is selection context, not execution evidence. A successful simulation is block- and state-specific and is not proof that a later mainnet transaction will succeed.
MANUAL_REVIEW means only that all mandatory checks completed against the
available evidence and no documented stop condition was detected. A person
must still inspect wallet details and current conditions. The label must never
be rendered or described as safe, approved, guaranteed, or ready to sign.
The only alternative result is STOP.