@honua/sdk-js
    Preparing search index...

    Module agent-safety

    Deterministic safety boundary for JSON-compatible agent plans.

    This module never invokes a model, tool, source, renderer, or job. Dry runs are side-effect free. Cryptographic operations and the atomic single-use replay store are delegated to host-provided callbacks so key custody and persistence stay outside the SDK.

    This entrypoint is part of the SDK's stable tier: symbols reachable from @honua/sdk-js/agent-safety are covered by the semver contract (see docs/decisions/agent-surface-stabilization.md and docs/agent-safety-threat-model.md).

    Classes

    HonuaAgentExecutionError
    HonuaAgentSafetyError

    Interfaces

    AgentApprovalConsumptionV1
    AgentApprovalRequestV1
    AgentApprovalUseConsumer
    AgentApprovalV1
    AgentApprovedStepV1
    AgentCitationV1
    AgentDryRunV1
    AgentEffectBudgetV1
    AgentEnvelopeSigner
    AgentEnvelopeVerifier
    AgentExecutionAuditSinkV1
    AgentExecutionCompletedAuditV1
    AgentExecutionContextV1
    AgentExecutionEvidenceV1
    AgentExecutionInputSnapshotV1
    AgentExecutionReceiptV1
    AgentExecutionStartedAuditV1
    AgentOperationExecutionResultV1
    AgentOperationExecutorV1
    AgentOperationInputV1
    AgentPlanPolicyV1
    AgentPlanStepV1
    AgentPlanV1
    AgentProvenanceV1
    AgentQueryPlanBindingV1
    AgentSafetyEvidenceProvenanceV1
    AgentSafetyEvidenceV1
    AgentSafetyOptions
    AgentSourceBindingV1
    AgentSourcePolicyV1
    AgentStepAuthorizationV1
    DeriveAgentSafetyEvidenceOptions
    ExecuteAgentPlanStepOptions
    ExecutedAgentPlanStepV1

    Type Aliases

    AgentDataMode
    AgentDigest
    AgentEffect
    AgentExecutionAuditV1
    AgentSafetyErrorCode
    AgentSafetyUnavailableFact

    Variables

    AGENT_APPROVAL_KIND
    AGENT_CONSUMPTION_KIND
    AGENT_DRY_RUN_KIND
    AGENT_EXECUTION_AUDIT_KIND
    AGENT_PLAN_KIND
    AGENT_RECEIPT_KIND
    AGENT_SAFETY_EVIDENCE_KIND
    AGENT_SAFETY_HARD_LIMITS
    AGENT_SAFETY_VERSION

    Functions

    deriveAgentSafetyEvidence
    digestAgentOperationInput
    dryRunAgentPlan
    executeAgentPlanStep
    issueAgentApproval
    issueAgentExecutionReceipt
    snapshotAgentExecutionInputs
    verifyAgentApproval
    verifyAgentExecutionReceipt
    verifyAgentSafetyEvidence
    verifyAgentStepAuthorization