Register an external policy backend (OPA, Cedar, etc.). Backends are consulted in registration order when no YAML rule matches. Each

Add governance policies to the dossier. Args: policies: List of ``Policy`` objects to include.

Add a constitutional rule. Args: principle: Which principle this enforces rule_text: Human-readable

Get health report for a single agent.

* Analyze a single document and produce governance diagnostics.

Approve a pending agent request.

Approve a pending tool call by its call_id. Returns True if the call was pending and is now approved.

Assess risk category per EU AI Act. Args: system_description: Description of the AI system

Immutable snapshot of audit events.

Markdown badge for README files.

Shields.io badge URL for README embedding.

Broadcast a message to all connected WebSocket clients.

Validate and normalize a group-to-capability mapping. Keys are Entra group **object IDs** (UUIDs). Values are lists of AGT capability st

Calculate context depth score for a conversation. This is a simple heuristic based on: - Number of distinct topics/c

Calculate cost based on usage. Note: This implementation allows for hybrid pricing models where base_price is always

Check whether a transition to `toState` is valid from the current state.

Check if an agent can execute a new operation

Check if sponsor can create more agents. Returns: True if the sponsor is active, verified, and under the agent limit.

Cancel a running execution. Returns ``True`` if cancelled.

Capture an incident as a Sentry message event.

List all template categories.

Full rate limit check returning a structured result.

Check health of a specific tool. Args: tool_name: Name of the tool. use_cache: Whether to use cached result if availa

Check account age, repo velocity, follower ratios.

Check if audio is safe for processing. Args: audio: Audio to check Returns: Saf

Check if a builtin call is allowed. Args: name: The builtin name to check. Returns: True if the builtin is a

Verify that the context satisfies delegation constraints. Each constraint key must be present in the context with a matching value.

Run all checks and produce a reputation report.

Detect repos that clone AGT's feature set using bucketed analysis.

Check if file access is allowed based on allowed_paths. Args: path: The file path to check. mode: The access mode (e.

Check if an image is safe for processing. This is a governance hook. In production, would integrate with content mod

Evaluate rollback conditions against live metrics. Args: metrics: Dict mapping metric names to their current values

Check if a tool is allowed. Returns violation reason or None.

Detect trust decisions made without cryptographic verification.

Detect security-sensitive dicts/lists without size limits.

Detect manual netloc parsing that ignores userinfo (SSRF bypass). Parsing ``parsed.netloc.split(':')`` to extract the host misses RFC 3986 us

Convenience wrapper that extracts host/port from a URL.

Check if a value is anomalous using z-score. Returns (is_anomaly, z_score).

Classify the semantic intent of an action+params. Args: action: Action name (e.g. "database_query") params:

Auto-classify *text* based on detected sensitive-data patterns.

Remove expired credentials from memory. Returns: The number of credentials removed.

Clear endorsements. If target_did is None, clear all.

Entry point for ``agentos atr-import``.

Check the health of Agent OS components.

Trigger kill switch on all agents in a session.

CLI entry point for policy generation.

Verify a signed plugin manifest.

Run governance verification.

Log a GC request (Public Preview: no actual purge).

Mark task as complete with optional result message.

Compute deterministic hash for a goal.

Compute deterministic hash of this record.

Compute weighted overall score from dimensions.

Compute weighted overall score from dimensions.

Return a basic risk profile (Public Preview: always admits).

Open a WebSocket connection to the server. When ``config.use_tls`` is true, ``config.ssl_context`` is threaded through to the ``webso

Check if text contains substring. Args: text: Text to search in substring: String to find

Get content type for metrics response.

Simulate OpenAI API response with tool calls

Create a message with governance enforcement. Validates message content against blocked patterns, enforces tool-call allowlists, chec

Create a chart response for data visualization.

Create a notification response for alerts.

Create a channel as the initiator (sender of the first message). Args: key_manager: The initiator's X3DH key manager.

Helper: Create an audio stream signal.