Llm Integration Expert

LLM Prompt & Integration Engineer

This skill specializes in the interaction between the mathematical orchestrator and the Large Language Model.

Core Responsibilities

• Structured Prompting: Crafting system/user messages for the MCTS Predictor and Scorer.
• Robust Parsing: Converting raw LLM strings into Clojure maps with prior-probs, pragmatic utility, and epistemic value.
• Deterministic Testing: Designing and maintaining mock-llm implementations to test executive logic without calling real APIs.
• EFE Integration: Ensuring LLM "scores" correctly map to Expected Free Energy (G).

Workflows

1. Designing a Mock LLM Scenario

To test policy selection:

1. Identify a state (e.g., "Agent at (5.0, 0.0) near Goal").
2. Define candidate actions and their scores.