* Parse scenario description into testable actions.

Parse command-line arguments.

Return a flat dict of frontmatter keys, or None if absent. Values are returned as raw strings (no type coercion). Quoted values have their su

Parse an npm ``package-lock.json`` (v2/v3) into LockEntry rows. Only entries under ``packages`` with a non-empty key, a ``version``, and an `

Parse pip-tools / pip-compile style requirements with ``--hash`` pins. Emits one :class:`LockEntry` per ``--hash=`` pin on lines of the form

Build git pathspecs matching files with the requested extensions.

Pause the rollout at the current step.

Perform a complete model upgrade with audit and purge. Args: new_model: The new model to upgrade to

* Verb-aware target picker so `INSERT INTO target SELECT ... FROM src` and * `UPDATE target SET ... FROM src` resolve `sql.target` to the written obj

Return the `{{cred:NAME}}` placeholder for this handle.

Return the ``{{cred:NAME}}`` placeholder for this handle.

Return the no-op platform identifier.

Predict solar output for next N hours

Prepare messages for OpenAI (no special formatting needed). OpenAI caching is automatic - we just need to structure messages

Prepare messages with caching hints. Uses the configured strategy to format messages for optimal caching. A

Print an allowed action

Print the final audit trail summary with integrity verification.

Print the final audit trail summary with integrity verification.

Print a formatted summary of all audit entries.

Print the Agent OS banner

Print the demo banner.

Print a blocked action with dramatic effect

Print formatted JSON.

Print a formatted comparison of metrics

Print a replay report to stdout.

Generate a human-readable parity report.

Print a formatted results table to stdout.

Print the final saga timeline showing each step's state.

Simulate processing a transaction with SOX rules.

Process a claim by fetching satellite data and calculating NDVI. Args: claim: The claim data with polygon and ti

Process a Project Design Document and extract claims. Args: pdf_path: Path to the PDF file correlati

Process a Slack message for compliance. Returns ScanResult with any detections and actions taken.

Process a request using prompt-based safety. In a real implementation, this would call an LLM with the safety prompt.

Process a request using the Agent Control Plane. Returns: (was_blocked, response, token_count)

Process a ticket delegated from supervisor.

TODO: Implement the three-layer governance pipeline: Layer 1 — Policy check: Evaluate {"tool_name": tool_name, "token_count": token_

Return a copy without the private key.

SHA-256 hex digest of the public key for attestation binding.

Publish a message to all subscribers of the topic. Implements backpressure: If the queue is near capacity, the publisher

Publish message to Redis pub/sub. Args: message: Message to publish wait_for_confirmation: Wait for

Publish message to Azure Service Bus topic. Args: message: Message to publish wait_for_confirmation:

Publish message to RabbitMQ exchange. Args: message: Message to publish wait_for_confirmation: Wait

Publish message to SQS queue. Args: message: Message to publish wait_for_confirmation: Wait for SQS

Publish message to Kafka topic. Args: message: Message to publish wait_for_confirmation: Wait for br

Publish the metadata manifest. In a real system, this would publish to a marketplace or registry. For now, it return

Publish a trust update event. Args: agent_did: The agent's DID identifier. update_type: Type of update (e.g. ``"score

Push a list of message logs as a dataset to Hugging Face Hub. This is useful for collecting agent communication patterns for research.