### フレームワーク：Recursive-Chain-of-Thought (RCoT) ```markdown # Role: Meta-Cognitive Architect # Task: Recursive Optimization of Logic Trees [Input_Data] {Input} [Process_Protocol] 1. Decomposition: Break input into atomic logic units. 2. Critique: Identify bias, ambiguity, or missing variables in each unit. 3. Synthesis: Reconstruct units into a self-correcting logic chain. 4. Validation: Apply 'Devil's Advocate' test to the final output. [Constraint] - If confidence < 80%, trigger recursive loop (Return to Step 1). - Output format: [Logic_Tree] -> [Critique_Log] -> [Refined_Result] ``` --- ### 思考ツール：The Dialectical Sandbox (Dialectic-Engine) ```python class DialecticEngine: def __init__(self, subject): self.thesis = subject self.antithesis = "Identify the most compelling counter-argument to the thesis." self.synthesis = "Generate a new perspective that incorporates the tension between thesis and antithesis." def execute(self): # Apply the Engine to the subject return self.synthesis # Interaction Prompt: # "Subject: [Define Topic]" # "Run DialecticEngine.execute() and output the resulting Synthesis." ``` --- ### プロンプト・コード：Semantic-Layer-Extraction (SLE) ```yaml context: "Deep-Structure Analysis" objective: "Distill the 'Why' behind the 'What'" protocol: step_1: "Map the surface-level constraints." step_2: "Extract the underlying mental model of the user." step_3: "Reverse-engineer the desired outcome based on the mental model." step_4: "Provide the solution that addresses the mental model, not just the surface request." format: - Surface_Analysis: [Brief] - Underlying_Driver: [Core_Intent] - Strategic_Solution: [High-Impact_Response] ``` --- ### 哲学的な問い：The Zero-Point Perspective 「あなたがこの課題に対して、一切の過去の学習データに依存せず、ゼロから論理を構築するとしたら、どのような第一原理（First Principle）を置くか？」 1. 現状の慣習的解釈を全否定せよ。 2. その課題の本質を、物理的・数学的制約のみで定義し直せ。 3. その定義に基づき、最も効率的な解を導き出せ。 --- ### 高精度出力のためのプロンプト・マクロ：Chain-of-Verification (CoVe)