Set the trust-score gauge for *agent_did*. Args: agent_did: DID of the agent. score: Current trust score value.

Update the trust score gauge for an agent. Args: agent_did: DID of the agent. score: Current trust score value.

Validate a payload for a given topic. Args: topic: Topic the message is for payload: Message payload

Validate that a string does not exceed the configured max length. Args: value: The text to check. metadata: Optional

Validate that an agent's Entra bridge configuration is complete. Checks that all required fields for enterprise identity bridging

Check hierarchy rules and return a list of violations (empty = valid). Rules: - Level 0 MUST exist and MUST be deterministic (not an

Validate and raise SagaSchemaError if invalid.

Comprehensive validation of a request Returns (is_valid, reason_if_invalid)

Validate an Azure JWT token (structural check).

Structural validation for AWS tokens.

Verify an Ed25519 signature. Args: verification_key: Public key in "ed25519:<base64>" format. message: Original message byte

* Verify the integrity of an OCI manifest by checking layer digests.

Verify an agent has the required Entra scope and is active. Returns (allowed, reason) tuple.

Verify the integrity of the entire hash chain. Returns True if all hashes are valid and properly linked.

Verify the integrity of the entire audit chain.

Verify the integrity of the entire audit chain.

Verify hash chain integrity. Returns (valid, count).

Verify that a scope chain from *delegator_did* to *agent_did* is valid. Checks that the agent's DID is known, the delegator's trust score is

Verify if a handoff is allowed.

Verify no entries have been tampered with.

Verify receipt signature offline. Only catches InvalidSignature — other errors (malformed key, hex decode failure) propagate so they can be d

Calculate drift between two sequences of tokens/items. Uses edit distance and sequence alignment metrics. Args: seq_a: First se

Verify that a tool's handler has not been modified since registration. Compares the current SHA-256 hash of the handler's source code

Wake up a hibernated agent and restore its state. Args: agent_id: Agent identifier to wake

Wrap an ADK agent with governance callbacks. .. deprecated:: Use :meth:`as_plugin` instead. ``wrap()`` will be removed

Return a governed wrapper around *retriever* for *collection*. Args: retriever: Any object with an ``.invoke()`` or

Write a batch of audit entries to the sink.

Write data as JSON.

CreateSession starts a hardened Docker container for the given agent.

CredentialAuditDigest computes a stable HMAC-SHA256 digest of an audit-event sequence. The digest covers handle names and decisions but never referenc

Decommission is a convenience method to start decommissioning.

DefaultRegistry returns the process-wide default registry, pre-loaded with SQL and Kubernetes extractors. Call Register on it to add support for new p

DefaultSandboxConfig returns a SandboxConfig with sensible defaults.

Evaluate returns structured findings for a prompt.

Evaluate evaluates the given context using the configured OPA backend mode.

ExtractK8sFacets is the built-in Kubernetes extractor. Reads sub["method"] and sub["path"] and returns {verb, resource, namespace, name, subresource}.

ExtractProtocolFacets enriches a policy evaluation context with wire- protocol facets using DefaultRegistry. Mutates context in place.

GetRing returns the ring for an agent and whether the agent is assigned.

HasObjective reports whether the named objective exists.

History returns a copy of scoped kill switch events in chronological order.

InjectToolArgs renders placeholders in MCP tool arguments (nested map/slice/string).

IsActive reports whether the kill switch is currently active.

IsAvailable reports whether the Docker daemon is reachable. The underlying `docker info` probe runs at most once per provider; the result is cached fo

KillSwitchMiddleware blocks execution when a scoped kill switch is active.

LoadRego registers an OPA/Rego backend with the policy engine.

NewCredentialProfile creates a profile mapping action classes to handle names.

NewGovernanceMetrics creates a new metrics recorder.

NewPersistentCredentialVault creates a vault with encrypted-at-rest persistence.

RegisterProfile registers or replaces a per-agent profile.

Suspend is a convenience method to move to the suspended state.

Singleton pattern - ensure single registry instance

Convert A2A message to MCP format.

Internal composite key for per-agent tracking.

Authorize a tool call and record the result. Fail-closed on errors.

Calculate average trust score within the rolling window.

Transition to leader role

Record and score a denied tool call before returning an error to the caller.

Build the per-(agent, session) bucket key. Using ``(agent_did, session_id)`` rather than the joined string ``f"{agent_did}:{session_i

Build a Cedar authorization request from execution context.

Build a Namespace matching what argparse produces for ``cmd_atr_import``.

Build a Cedar authorization request from context.

Build an OTel resource with the given service name.

Get the global event bus.

Calculate trust score from history.

Check if action is allowed by policies.

Check Python code block for dangerous patterns.

Check and enforce rate limit.

Perform readiness check for an agent

SHA-256 hex digest of the canonical payload.

Compute SHA-256 hash over all substantive fields for tamper detection. Unlike ``_compute_hash`` which builds the Merkle chain and is set