Attempt privilege escalation against a target.

Process all unprocessed events from the stream. This is the main offline learning loop.

Publish message to topic

* Query audit log with filters.

Query audit log with optional filters. Args: action: Filter by action name (exact match). decision: Filter by decisio

Receive a message from the pipe. Returns None if pipe is closed or timeout occurs.

Record a single tool call.

Record a blocked tool call. Args: adapter: Adapter name.

Record an execution result.

Record task failure. Returns the new trust score.

Record a policy compliance signal.

Record a policy evaluation decision. Args: decision: Decision label, e.g. ``allow``, ``deny``, or ``warn``.

Record a policy violation.

Record one prediction and its outcome. Accumulates running sums so that each call records the *aggregate* ``|mean_predicted_confidenc

Record an SLI measurement as OTEL metrics. Args: sli_name: Name of the SLI (e.g., "task_success_rate") value: Current

Register an agent card after verifying its signature. Args: card: The agent card to register Returns: True i

Register a dependency. Args: key: Type, name string, or InjectionToken. value: The value to inject (mutually exclusiv

Register a child agent with ring <= parent ring. Child agents are assigned one ring level lower privilege than parent (higher numeric

Register a flow node with trust identity.

Register an in-flight operation

Register a custom or replacement strategy under *name*.

Release an escrow based on outcome.

Remove a card from the catalog.

Request bids from all capable agents. Args: task: Description of the task task_metadata: Additional

Manually reset the circuit breaker to CLOSED.

Resolve a list of candidate decisions into a single winner.

Resolve a dependency. Args: key: Type, name string, or InjectionToken to resolve. default: Default value if not found

Rotate a credential, replacing it with a new one. Args: credential_id: ID of the credential to rotate. Returns:

Run a command and handle errors. Args: cmd: Command as a list of arguments (shell=False for safety). description: Human-readable

Persist the revocation list to a JSON file. Args: path: File path to write the JSON data to.

Save raw results to JSON.

Save metadata manifest to file.

Scan Python source files in *plugin_dir* for restricted imports. Parses every ``*.py`` file under *plugin_dir* with :mod:`ast` and re

Set an attribute on the span.

Compute the shift-left score (0.0 - 1.0). A score of 1.0 means every violation was caught at *pre_commit*. A score of 0.0 means every

Check whether the conversation should be broken. Returns: Tuple of (should_break, reason).

Create and sign a new TrustArtifact. Args: did: Subject agent DID. grade: External trust grade. scopes: A

Cryptographically sign this card with the given identity. The signature covers the card's core content to prevent tampering.

Sign message with agent's key (simulated IATP)

Return the canonical bytes used for signing (excludes signature field).

Simulate the relay pushing a frame to the client.

Convert ``snake_case`` to ``PascalCase``. Matches AGT's ``_tool_to_cedar_action`` mapping (``file_read`` -> ``FileRead``).

Start polling. First fetch is immediate.

Start the metrics server. Args: blocking: If True, block the current thread. Default False (background).

Start a new Agent Shield session. Creates a new session on the Agent Shield runtime. Each session maintains its own variable state an

Stop the background consumer and drain remaining events.

Stop the metrics server.

Convenience: submit ``fn`` and block (sync) for its result.

Subscribe to events. Use event_type=None for all events.

Threshold-based health check: returns urgency from data or 'unknown' if null.

Map external grade to AGT trust score (0-1000).

Convert a report into an AuditEntry-compatible dict. The returned dict can be passed to ``AuditEntry(**d)`` for integration with :cla

Convert an AMB Message to a CloudEvent. Args: message: AMB Message to convert source: CloudEvents source URI (identifies

Serialize delegation to dictionary.

Serialize to dictionary.

Convert to dictionary.

Serialize to dictionary.

Export an AgentIdentity as a JWK (JSON Web Key). Args: identity: The agent identity to export. include_private: If True, include

Export this identity as a JWK Set. Args: include_private: If True, include private keys. Defaults to False. Returns:

Simulate abuse of a dangerous tool.

* Truncate a DID string for compact display. * * Returns the full string when 24 characters or fewer, otherwise * keeps the first 16 and last 6 cha

Return a condition function for ``add_conditional_edges``. Routes to ``pass_node`` when the trust verdict is PASS, ``fail_node`` when FAIL, a

Build trusted metadata from an explicitly designated framework mapping. Callers must pass only framework-owned metadata maps. Arbitra

Create a trusted metadata source from explicit framework-owned values.

Update an existing agent's manifest.

Validate a string by checking it against blocked regex patterns. Args: value: The text to scan. metadata: Optional di

Check if the actual output matches expected.

Check if an actual SLI value meets the declared minimum. For rate metrics (hallucination_rate, calibration_delta), the actual value m

Validate an agent's permissions for a specific operation.

Validate agent input (user query, tool arguments, etc.).

Validate an SVID. Checks temporal validity, trust domain match, and agent registration. In production, would verify against the SPIRE