Complete this trace.

Load configuration from AGENT_OS_* environment variables.

Load a policy from a YAML file, resolving ``extends`` parents. Recursively loads parent policies referenced by ``extends``, merging r

Look up an action's reversibility entry.

Mock implementations of script functions for testing

Get agent metadata.

Get processor for given format.

Get query statistics.

Return a normalised trust score (0.0-1.0) from AVP EigenTrust. Calls GET /v1/reputation/{did} and returns the ``score`` field. Return

Execute a tool call through the governance layer. Returns result or None.

Grant a capability to an agent. Creates a ``CapabilityGrant``, adds it to the agent's scope, and tracks it by grantor for bulk revoca

Guard common OpenAI-style async client create methods by duck typing.

Generate overall health summary.

Check if an agent is currently quarantined.

Issue a new SVID for this identity. In production, this would request an SVID from the SPIRE server. Args: ttl_hours: Ho

Return manifests for all installed plugins. On-disk manifests are re-verified against ``self._trusted_keys`` so a tampered or unsigne

Merge a chain of PolicyDocuments into a flat, priority-sorted rule list. Rules from all levels are collected. A child rule with ``override=Tr

Get current node positions.

Send event to all matching webhooks synchronously.

Parse a semantic version string into a tuple. Args: version: Version string (e.g., "1.2.3"). Returns: Tuple of (major, minor

Process PDF content.

Record a good or bad event against the budget. Events are stored in a bounded deque. When ``max_events`` is reached, the oldest even

Record a hallucination detected by cmvk and slash reputation. Args: agent_id: Agent that hallucinated severi

Register an agent and create SPIFFE identity. Returns the existing identity if already registered. Args: agent_did: The

Record an audit entry and return ``entry_id`` and ``chain_hash``.

Execute query and emit telemetry. No reflection or evolution - just execution. Args: query: User's query

Sign a message with an Ed25519 private key. Args: private_key_bytes: Raw 32-byte Ed25519 private key. message: Bytes to sign

Start the scaling manager

Begin the rollout, moving to the first step. Sets state to ``CANARY``, ``current_step_index`` to 0, and records a ``step_start`` even

Auto-tune weights for document sections based on content analysis. Uses structure-aware indexing to assign tiers and calcula

Update the trust score, respecting ceiling if set.

* Update status bar based on enabled state

Validate an AI-generated plan against all constraint rules. Args: plan: Dictionary containing the action plan

Validate request against risk policies

Validate a session token for a specific agent. Args: agent_id: Agent identifier expected to own the session. session_

Run governance verification across all ASI controls.

Verify client identity before allowing tool access.

Verify the entire delegation chain from *identity* to the root. Walks ``parent_did`` links, checking at each level that: 1. The paren

Verify hash-chain integrity and Ed25519 signatures for an ordered receipt list. Returns a list of error strings; empty means the chain is fully v

Assign places an agent in the specified privilege ring.

ExecuteWithGovernance evaluates the action through the governance pipeline.

RecordFailure decreases an agent's trust score with asymmetric penalty.

Sign signs data with the agent's private key.

Unique agent id per test invocation, capped to the Azure name regex.

Check whether *tool_name* is permitted by the active policy. Evaluation order: 1. If *tool_name* is in ``_blocked_tools`` → deny.

Set command execution policy. Call once from CLI entry point.

Down-grade volume/activity signals for accounts with organic credibility. Dampening is skipped entirely when abuse-pattern signals (credential

Build a common denied policy-check result.

Fallback SQL check when sqlglot is not available. Uses regex pattern matching. Rules are driven by *config*; when *config* is ``None`` a

SHA-256 hash of a function's compiled code object. Hashes the marshalled :class:`code` object (which covers ``co_code``, ``co_consts``, ``co_

Build an httpx.AsyncClient.get mock that dispatches by URL. Responses include a Request so `raise_for_status()` works.

Create a mock CrewAI agent with tools.

Return a detector with frequency/entropy disabled for determinism.

Create a mock CrewAI crew.

Create a mock tool result (ToolMessage).

Create a mock urllib response.

Recalculate trust score from recent signals. Score is calculated as weighted sum of dimension scores, where each dimension i

Release a bounded-semaphore slot if this context currently holds one.

Simulate on_llm_start → on_llm_end (or on_llm_error).

Simulate on_tool_start → on_tool_end (or on_tool_error).

Reject ``agent_id`` values that would be unsafe to interpolate into Docker / kubectl command-lines and label values. Returns the validated id

Write a valid policy YAML, optionally overriding top-level keys.

Add a capability to the agent.

Add a scoring dimension to the rubric. Args: dimension: Name of the dimension (e.g., "correctness", "tone")

Add context labelled as *hot*.

Generate a dedup fingerprint from alert fields. Args: alert: The alert to fingerprint. fields: Tuple of Alert attribute names to

Resolve a tool name to its canonical form. Checks exact aliases first, then regex patterns. Returns the original name (lowercased) if

Classify and enforce — raises PolicyDenied if intent is denied. Args: action: Action name params: Action par

Evaluate all applicable gates for a proposed promotion. Args: agent_id: Agent identifier. current: Current maturity l

Check pyproject.toml for loose version constraints. Uses ``tomllib`` (Python 3.11+) so the following layouts are all covered, includi

Check if a tool call is allowed by policy. Runs all governance checks: allowlist, call count, content, and semantic intent classifica

Compile ATR rules into one PolicyDocument YAML per category. Returns a manifest dict summarising what was written. Reuses the PR #908 conver