What problem does it solve? Autonomous decision workflows need explicit owner guidance before approving, denying, or escalating changes to controls and policies. This Skill standardizes how agents fetch the markdown guidance pods entity owners attach to those entities, and enforces a 0.70 confidence cap when no guidance exists. ## Core Features & Use Cases - Pod Fetch Contract: Defines the GET endpoints for listing all llm_context_rule pods on an entity and retrieving a specific pod's full markdown content by key. - Parent-Walk Pattern: When the target is a policy or policy_exception, also fetches the parent control's pod, preferring policy-level guidance for policy-specific decisions and control-level guidance for broader rules. - Confidence Cap Precondition: Establishes that missing pods on both target and parent cap decision confidence at 0.70, forcing advisory-only mode with a comment for human review. - Use Case: During an autonomous ticket approval, the agent fetches the policy's pod, finds guidance like "auto-deny any policy that removes the internet_facing index", and applies it to the decision. ## Quick Start Fetch the LLM context rule pods for this policy and its parent control, then tell me whether owner guidance exists or if we are capped at 0.70 confidence.
