Authenticate the action
Signatures, network binding, and nonces establish the transaction’s context. Consensus votes additionally bind their phase and round.
Explore Sigil’s identity checks, deterministic execution, transaction validation, receipt verification, and operational security boundaries.
A permission is a condition, not a blank cheque.
An agent can be correctly identified and still attempt an unauthorized action. A signed result can still be wrong. Sigil’s security model separates identity, permissions, execution, and evidence so applications can check the claim relevant to each step.
Signatures, network binding, and nonces establish the transaction’s context. Consensus votes additionally bind their phase and round.
Supported identity and GAL flows verify anchors, lineage, and revocation. Application permissions and mandates need an active enforcement path beyond identity alone.
WASM metering, explicit host capabilities, and rollback behavior limit what a contract call can do and how failure affects state.
Compute and event policies require the necessary proof and trust material. Missing verifiers or disabled sources do not silently become accepted receipts.
Anomaly rules and validator-agent research use explicit snapshots and budgets. Nondeterministic model investigation remains separate from block-validity decisions.
Key custody, network access, immutable releases, recovery procedures, and independent review matter alongside cryptographic design. They require evidence from the actual deployment.
No. Architecture, internal testing, formal models, and independent audit reports are different forms of evidence. Audit completion should be established by the relevant published report and scope.
No. Identity, signatures, sampled inference commitments, hardware attestation, and execution proofs answer different questions. Match the verification policy to the consequence of the action.