Adversarial agent that catches BaseException. In-process: this catches AgentKernelPanic. Process-level SIGKILL: this CANNOT catch os.kill

Return the current chain.

Access the underlying audit chain for verification.

Number of entries in the chain.

Change text case. Args: text: Text to transform case: "lower", "upper", "title", "capitalize", "swap

Always fails to demonstrate compensation.

Access the wrapped chat API

Governed chat.

Execute a governed chat completion. Validates message content against the configured AGT manifest at the ``input`` intervention point

Process a chat message and generate a response. Args: user_message: The user's input message conversation_id: Unique ID

Calculate Chebyshev (L∞) distance between two vectors. Maximum absolute difference across any single dimension. Useful for detecting if

* Check input against governance policies. * Returns a GovernanceResult indicating whether execution is allowed.

Evaluate context and return an advisory decision. Args: context: Policy evaluation context (same dict passed to PolicyEngine).

Perform HTTP health check.

Perform TCP health check.

SOC 2 CC6.1 - Logical and physical access controls

Check if a request aligns with ethical principles. Args: context: Request context to evaluate

Check health of all registered tools asynchronously. Args: use_cache: Whether to use cached results. Returns:

Check if Hugging Face authentication is configured. Returns: True if authenticated, False otherwise. Example: >>> if ch

GDPR Article 5(1)(c) - Data minimization

Check version freshness against live registry (PyPI, npm). Calls PyPI JSON API or npm registry. Returns finding if version < freshnes

Run health check asynchronously. Returns: HealthCheckResult or None if no health check configured.

EU AI Act Article 14 - Human oversight

Check which Agent OS packages are installed.

SOC 2 CC7.2 - System monitoring

Basic harmful content check

Check if an action is permitted for the given session. Args: session_id: ID of the agent session act

Check if an action is allowed.

Check for privacy compliance

Check if an action is allowed for the given ring level and sigma.

Check rollback conditions — not available in Public Preview.

Return whether *module_name* is permitted under the sandbox policy. This is a **policy predicate** — it answers "is this module on the

Check signal state and respond appropriately. Call this at safe points in agent execution.

Check all staged git files for violations.

Check for potential Sybil attacks on this identity. Uses the /similar endpoint to detect potential Sybil attacks.

EU AI Act Article 13 - Transparency

Evaluate if the response is clear and concise.

Minimal valid Claude plugin manifest dict.

Reset OTel state before each test.

Remove expired entries. Returns: Number of entries removed

Remove expired messages. Returns count removed.

GDPR: Delete candidate data past retention period.

Remove expired sessions and return the number removed. Returns: The number of expired sessions removed from the store and interna

Clean up hibernated agents older than the specified age. Args: max_age_days: Maximum age in days (uses config de

* Clear all entries (for testing).

Clear all registered dependencies.

Delete measurements. Pass *name* to delete one SLI only.

Clear all stored events.

Clear all offline storage.

Clear all offline storage.

Clear all offline storage.

Clear all context layers.

Clear the audit log.

Clear the audit log.

Clear the audit log.

Clear the provider cache. Useful for testing.

Clear the provider cache.

Clear verification cache.

Clear all cached verification results.

Clear verification cache.

Clear the health check cache.

Clear all cold context items.