Security model
How scoped authority, confidential ingress, protected execution and Standard Proof govern sensitive work.
Standard combines execution integrity, scoped authority, and confidentiality for managed confidential execution. This page defines the protection boundaries and the assumptions behind Standard Proof.
Zero-trust principles
Standard separates requesting work from authorizing the use of inputs, execution of a function and release of results. Each operation must satisfy the applicable owner permissions, accepted workload and execution requirements.
The receiving integration verifies Standard Proof against the expected operation and independently accepted references before accepting the result. Your workflow then evaluates the outcome and enforces permission for the next action, whether the same agent continues or another agent takes over.
These checks rely on explicitly accepted hardware and attestation roots, approved software, secure keys and correctly configured verifiers.
Customer-defined checks
Customers implement required calculations, validations and failure behavior inside the approved function. Establish its accepted workload revision outside the requesting agent’s authority. The caller requests the function with authorized inputs; admission checks the workload and scope, and rejects substitutions.
The function’s code determines its business result. Customer tests establish its intended behavior; Standard Proof binds the recorded execution to the accepted workload and inputs. A function can return a permitted report containing failed checks. The receiving integration verifies the evidence, evaluates that report and applies its own policy before a dependent action.
Restricted Wasm execution
The approved function runs as a Wasm module with declared input and output interfaces and resource limits. It reads only the inputs the operation declares, writes only declared output candidates and observes cancellation. It has no general shell, filesystem, subprocess or network interface, and it does not inherit the requesting agent’s filesystem, shell or credentials.
Memory, fuel, duration and input and output sizes are bounded by the accepted limits. Ingress’s scoped permission to use a source credential is separate from the workload’s interfaces: connected-source retrieval happens in confidential ingress, and the function receives only validated inputs. Output candidates leave only as the release rules permit. See Write a protected function for the interface.
Execution integrity
Runtime attestation identifies the accepted execution environment. Execution evidence binds the relevant workload, authority, inputs, and result within that environment. A recipient verifies those bindings against references it accepts independently of the producing agent.
Signed evidence is tamper-evident: changing signed contents must invalidate their verification. The strength of a claim depends on the evidence source, its binding to the result, and the accepted trust roots. A signature on an untrusted statement alone does not establish that the statement is true.