Verify checks the integrity of the entire hash chain. After rollover eviction, the surviving head's PreviousHash is checked against the seam hash reco

Pre-execution governance check for a tool call. Returns None to allow execution, or a dict with error info to block it. Ord

Create a task on a peer agent via the A2A protocol.

Get or create budget for an agent.

Issue a new credential for an agent. Args: agent_did: The agent's DID. capabilities: Scoped capabilities (subset of a

Add or update an entry in cache.

Read a file.

Validate and normalize URL.

Register an endorsement for a target agent.

* Add a context entry to the store and compute its integrity hash.

Define the 'Physics' of the agent using allow-list approach. This implements "Scale by Subtraction" - by defining what IS allowed,

Benchmark a synchronous function.

Remove all tools from the registry.

Get the current aggregated value within the window.

Install a plugin from the registry. Steps: 1. Resolve manifest from registry. 2. Verify Ed25519 signature when ``veri

Govern a Bedrock ``invoke_agent`` call. Enforces: 1. Rate limiting per agent ARN. 2. Blocked-pattern and PII scanning on ``in

Add an agent to this session.

Record a protocol response and finish the span.

Verify the card's signature is valid. Verification authority precedence (highest first): 1. Explicit ``identity`` -- author

Check if a version matches a constraint. Supports constraints like: - "1.0.0" - exact match - ">=1.0.0" - greater than or equal - ">1

String returns a stable string representation for logging and map keys.

Create a mock kernel suitable for most tests.

Run an async coroutine synchronously.

Write *content* to a YAML file and return its Path.

Main CLI entry point.

CLI entry point.

Create a mock kernel for testing without Agent Shield SDK. All validations will return allowed=True. Useful for unit tests and develo

* Register an agent in the dependency graph.

Add an agent to the revocation list. Args: agent_did: DID of the agent to revoke. reason: Human-readable reason for r

Send alert to all matching channels.

Try to acquire budget for an agent. Args: agent_id: Agent identifier whose budget should be decremented. Returns:

* Update status bar based on enabled state

Validate a definition and return list of errors (empty = valid).

Create a sponsorship record (Public Preview: always succeeds, no bonding).

NewAuditLogger creates an empty AuditLogger.

Create a minimal :class:`AuditEntry` for testing.

Return a minimal valid saga definition dict.

Add a delegation to the chain. Args: delegatee: The agent receiving the delegation capabilities: Capabilities being d

Remove all registered backends. Primarily for testing.

Helper to create a sample tool spec.

Encrypt a message with the current sending chain. Args: plaintext: Message bytes to encrypt. associated_data: Additio

Issue a new credential and store it in the manager. Args: agent_did: The agent's DID to issue the credential for. cap

Parse static tool definitions from config for tests/offline scans. Static tool definitions are optional in most real MCP configs. Live stdio

Evaluate governance policy before agent execution.

Update component scores and recalculate the total. Args: identity: Identity verification score (0–100). behavior: Beh

Create a mock PydanticAI Agent.

Create a mock OpenAI client with all required sub-objects.

* Declare a dependency: `fromAgent` depends on `toAgent`. * Creates a circuit breaker for this link.

Add or update a multi-agent policy.

Commit a session's Summary Hash.

Emit an event to all matching subscribers.

Export metrics in Prometheus format.

Retrieve a tool specification by name and optional version constraint. Args: name: The tool name version: Version con

Invoke a tool with trust verification. Args: tool_name: Name of tool to invoke arguments: Tool arguments

Register a plugin manifest. Args: manifest: The manifest to register. Raises: MarketplaceError: If the exact

Try to acquire a slot. Returns (acquired, reason).

Verify a peer agent's trustworthiness. Args: peer_card: The peer's agent card required_capabilities: Capabilities the

Wrap a PydanticAI Agent with governance. .. deprecated:: Use :meth:`as_capability` with ``capabilities=`` instead for

Create a handler from a kernel with the given policy.

Approve a declared intent, allowing execution to begin. Args: intent_id: The intent to approve. Returns: The

Convenience factory.

Fire all registered callbacks for an event type.

Emit an event to the stream.

Execute a callable function. Args: func: The callable function to execute. args: Dictionary of arguments to p

Complete this span.

Create a lightweight governance gate. Args: allow: Actions to allow (allowlist). If set, only these actions pass. deny: Actions t

Parse and register a policy from a YAML string. Args: yaml_content: Raw YAML policy definition. Returns: The

Register an agent in the fleet.

Assert callable either succeeds or raises an acceptable exception.

Execute an approved request

Retrieve a bundle by name and version.

The critical choke point. No tool executes without passing this gate. This is the "Hypervisor" pattern - intercepts tool calls BEFOR

Load a policy into the engine. Args: policy: Policy instance to register. Replaces any existing policy with the s

Parse sql_ctx["query"] and return verb, target, tables and functions. Uses sqlglot for AST level parsing. Falls back to UNKNOWN if sqlglot is

Create an X3DH key manager with a fresh Ed25519 identity.

Return a mock boto3 bedrock-agent-runtime client.

Add a document to the store. Args: document: The document to add Returns: The d

Try to consume one token. Returns ``True`` if the call is allowed.