Upload experiment results/logs to Hugging Face Hub. Args: results_path: Path to local results file (JSON, CSV, etc.) repo_id

Upload experiment logs or benchmark results to Hugging Face Hub. This function uploads a single file (typically JSON results from experiment

Upload experiment results to Hugging Face Hub. Args: file_path: Local path to the experiment results file (JSON/JSONL)

Upload experiment results to Hugging Face Hub. Args: results_path: Local path to the results file (JSON or folder). p

Upload a folder to Hugging Face Hub. Args: folder_path: Local folder path. path_in_repo: Target path within the repos

Upload the dataset to Hugging Face Hub. Args: repo_id: The repository ID on Hugging Face Hub dry_run: If True, only crea

Track the size of a container element via ResizeObserver.

Hook: manages scan rotation, pause/resume, and reduced motion detection.

Use Case 1: Data Analysis Pipeline An AI agent that analyzes customer data but must: - Only read from approved data sources - Ne

Use Case 2: Content Moderation Agent An AI agent that moderates user-generated content: - Can read content submissions - Can fla

Use Case 3: Testing New Agent in Shadow Mode Test a new autonomous agent in shadow mode before production: - Agent thinks it's execu

Use Case 4: Multi-Tenant SaaS Application Multiple customer tenants, each with their own agent: - Isolated permissions per tenant

Validate definition and return list of error messages (empty = valid). Performs both JSON schema validation and semantic checks: - Re

Validate if the raw signal has required fields. Args: raw_signal: Raw signal data to validate

Validate file change signal has required fields.

Validate log stream signal has required fields.

Validate audio signal has required fields.

validateOPAURL rejects URLs with non-HTTP schemes or known SSRF targets.

* Validate that policies can be applied together (no conflicts).

Check for dangerous code patterns.

Validate compatibility between this agent and another agent. Checks if this agent can work with another agent based on their

Quick validation against financial principles.

Always allow input text.

Validate plugin manifest files against the PluginManifest schema. Returns 0 if all files are valid, 1 if any fail.

Quick validation against medical principles.

Validate task outcome using SCAK. Replays the flight recorder log against the Control Plane to deterministically ver

Always allow output text, passing it through unchanged.

Validate parameters against tool schema. Returns: {"valid": bool, "errors": List[str]}

Check for path traversal attacks.

Return a list of guardrail violations. Empty list means the proposal is safe.

AST-based Python code validation. Uses ``ast.parse`` to walk the tree and applies heuristic regex checks on the raw source for patter

Validate if request maps to a defined capability. Implements CapabilityValidatorInterface.validate_request()

Quick validation against safety principles. Args: output: Text to validate context: Optional context Return

Check for destructive SQL.

Validate launch fields for local stdio MCP configs without launching them. Static-only scans should not execute configured commands, but they sho

Structural validation for GCP tokens.

Always allow tool calls.

Validate a tool call through Agent Shield Stages 2 and 3. Stage 2 (state validation) checks whether the tool call is allowed given th

Always allow tool results.

Validate a tool result through Agent Shield Stage 4. Stage 4 screens the raw tool output for leaked secrets, blocked content, and app

Validate semantic version format

Validate that weights sum to approximately 1.0.

Calculate variance between numeric claims

Verify the entire scope chain.

Check that the content hash and HMAC signature are valid. Args: secret_key: The HMAC secret used when the entry was signed.

Verify previous hash pointers across full receipt list.

Verify a chain of exported receipts. Returns (exit_code, per_receipt_results). exit_code 0 = valid, 1 = errors.

Verify chart for contradictions and errors.

Verify findings across multiple models. Returns: (agreement_score, verified_findings)

Recompute hash and compare to stored value without mutation.

Verify that this entry's stored hash matches a fresh computation. Returns: ``True`` if ``entry_hash`` equals ``compute_hash()``.

Verify customer identity before sensitive actions.

Check all governance toolkit packages and critical dependencies.

Verify a peer agent trustworthiness.

Verify a peer agent's identity and capabilities.

Verify peer agent before task interaction. Verification uses the AI Card attached to the peer's A2A card. Args: peer_did

Verify a peer before inter-agent communication.

Perform mathematical verification using cmvk. This is where "the Math decides, not the AI." Args:

Verify signed reasoning envelope and governance bindings.

Verify response consistency across models. Returns: (agreement_score, model_opinions) In production, this calls actu

Verify a signature against this identity's public key. Args: data: The original data that was signed signature: The s

Verify a signature against this agent's public key. All verification failures are logged at DEBUG. Bad signatures are an expected, at

Verify identity via the underlying AI Card signature.

Calculate violation rate as a percentage string.

Return all violation records for a given rule name.

Block until the API health endpoint responds.

Wake up a hibernated agent by restoring its state from disk. Args: agent_id: Unique identifier for the agent to

Watch audit logs for policy violations and suspicious patterns.

Watch execution logs and detect violations. Returns list of new violations detected.

Calculate weighted Euclidean distance between two vectors. Allows certain dimensions to contribute more to the overall distance. Args:

Run a what-if trace comparison with modified span outputs. Replays the trace with the given overrides applied, returning a ReplayResu

Get all sponsors for a given agent.