Simulate on_retriever_start → on_retriever_end (or on_retriever_error).

Validate *value* is safe to interpolate into a Docker resource name. Raises ``ValueError`` if *value* contains characters that could cause na

Add a test case to the golden dataset. This is the primary way engineers "write code" in Eval-DD.

Add a constraint edge and re-sort by descending priority.

Add a signal and update score.

* Analyze code for policy violations

Create a ``GovernanceMessageHook`` for non-invasive integration. The hook governs ``messages.create()`` calls without wrapping or pro

Return a :class:`GovernancePlugin` backed by this kernel. The plugin implements all 12 ADK ``BasePlugin`` lifecycle hooks and delegat

Pre-tool hook: validate tool call against governance policy. Routes the call through the AGT 5.0 ACS engine at the ``pre_tool_call``

Calculate distance between two vectors using specified metric. This is the primary entry point for distance calculations. Args:

Check if a session can access a given path. Args: session_id: Session requesting access. path: Path being accessed.

Check Cargo.toml for unpinned versions. Uses ``tomllib`` (Python 3.11+) so the following layouts are all covered, including ones the

Check a package.json for unpinned versions. Catches three failure modes: * registry-bypassing protocol specifiers (``gi

Validate policy YAML files against the policy schema. Parses each file, runs JSON Schema and structural validation, and reports errors with f

* Configure provider-specific settings. * * @param settings - Key-value settings for the provider.

Try to consume tokens. Returns True if allowed.

Create a new human sponsor. Args: email: Verified email address of the sponsor. name: Human-readable name.

Create a complete governance middleware stack for a MAF agent. Assembles and returns an ordered list of middleware instances ready to pass di

Start a new run. Args: name: Run name run_type: Run type (chain, llm, tool, retriever) inputs: Optional i

Create sample agents for demonstration. Returns: List of sample agent listings

Create a decorator that applies governance checks to tool functions. .. deprecated:: Tool governance is now handled automatically

Detect the document type based on content analysis. Args: document: The document to analyze

Detect potential policy violations in proposed wisdom update. This is the core of the "Policy Review" from the problem state

Evaluate a Cedar authorization request. Args: action: Cedar action string (e.g., ``Action::"ReadData"``) con

Execute action with kernel governance.

Assemble all ingested artifacts into an ``AnnexIVDocument``. Returns: A fully populated ``AnnexIVDocument`` with auto-generated

Deserialize a card from a dictionary. Performs safe access on all keys so missing values fall back to sensible defaults. Arg

Get an agent by fingerprint.

* Get agent by ID.

Return all recorded audit events as a list of dicts. Each entry contains ``type``, ``timestamp`` (ISO-8601), and ``data`` keys. The

Return the full audit trail.

Get effective MCP server policy for an organization. Org policies inherit from the base (enterprise) MCP policy. Org can add to allow

Query audit entries with optional filters.

Get the most recent entries for a specific agent.

Retrieve the handle for a specific agent.

Get current circuit breaker metrics

Get or create trust score for an agent.

Retrieve conversation transcript entries. Thread-safe.

Return current trust score (0–1000) for an agent.

Guard a duck-typed MCP tool provider/server ``call_tool`` method.

Check if this agent has a specific capability. Supports wildcard matching.

Inspect a Streamable HTTP or legacy SSE MCP server.

Check if escrow is in an active state.

Return ``True`` if the circuit-breaker is tripped for this agent.

Return whether the wrapper has received SIGKILL.

Send real OS SIGKILL -- truly non-catchable. On Unix: ``os.kill(pid, signal.SIGKILL)`` On Windows: ``TerminateProcess`` via ``proc

List all registered tools.

Load an SLO template by name. Args: name: Template name (e.g. "coding-agent") specs_dir: Optional custom specs directory pat

Add entry to audit log.

Make an HTTP request and handle errors gracefully.

Return *True* if this rule matches the given endpoint.

A test tool.

Publish an event to all current subscribers. Fault-isolated per listener.

Return all measurements for *name* with timestamp >= *since*.

Record a new audit entry with hash chain hash.

Record a task cost.

Record a request.

Register a substitute agent for a session.

Restore VFS to a previous snapshot.

Revoke this identity permanently.

Scale an agent type to a specific number of replicas

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

Set path-level permissions.

Return the canonical bytes used for signing (excludes signature field). Uses JSON with ``sort_keys=True``, ASCII-safe escaping, no wh

Start all registered agents in dependency order. Returns: Summary of startup results

Start a new span within a trace. Args: trace_id: Trace ID operation_name: Operation name

Stop the scaling manager

Stop all agents gracefully. Returns: Summary of shutdown results

JSON attestation for machine consumption.

Export this identity as a JWK (JSON Web Key). Args: include_private: If True, include the private key. Defaults to False.

Return a v4 ``PolicyCheckResult`` populated from this result. Only importable when ``agent_os`` is installed. Hosts that have not yet

Current token count after refill.

Return the canonical trust-level label for a numeric score (0-1000). The mapping mirrors the tiers defined in ``agentmesh.constants`` and is

Remove a tool from the registry. Args: name: The tool name version: Specific version to remove, or None to remove all