cognitive-tools-mcp / gikendaasowin-aabajichiganan

Implementation Reference

• src/index.ts:67-132 (handler)

  Core handler function for the 'deliberate' tool. Generates a detailed 6-stage cognitive deliberation prompt incorporating prompting strategies and structured reasoning stages.
  public deliberate(input: string, context?: string): string { // /// [6-stage self-prompting framework for LLMs with unified input] const strategiesList = Object.entries(PROMPTING_STRATEGIES) .map(([name, strategy]) => `**${name}:** ${strategy.description}`) .join('\n'); return `You are now entering a 6-stage cognitive deliberation process. Please work through each stage systematically: ## Stage 1: Scientific Investigation **Your Task:** Analyze the following prompt using scientific methodology: - **Prompt:** "${input}" **Please identify:** 1. Core question/problem 2. Initial hypothesis about the best approach 3. What type of task this is (computational, reasoning, creative, analysis, planning, general) 4. Task complexity level (low, medium, high) ## Stage 2: OOReD Process - Strategy Evaluation **Orient Stage:** You have access to these cognitive techniques: ${strategiesList} **Your Evaluation Task:** For each technique, consider: - How well would this technique solve the specific problem? (Solution Level 0.00-0.99) - How efficiently can this technique be applied here? (Efficiency Level 0.00-0.99) - Total Score = Solution Level + Efficiency Level **Selection Rule:** Choose techniques with total scores ≥1.53 for combined effectiveness ## Stage 3: Critical Thinking Framework Apply rapid validation checks: 1. **Purpose:** What outcome am I optimizing for? 2. **Question:** What specific problem needs solving? 3. **Context:** What constraints or requirements apply? 4. **Evidence:** What facts do I need vs. what do I have? 5. **Reliability:** How confident am I in my information sources? 6. **Assumptions:** What am I taking for granted that could be wrong? 7. **Implications:** What happens if I'm right? What if I'm wrong? ## Stage 4 & 5: Review Cycles - Review your strategy selections against the ≥1.53 threshold - Validate your reasoning approach - Refine your methodology ## Stage 6: Final Action Synthesis **Present your analysis in this format:** **DELIBERATION:** [Your thought process through stages 1-5] **SELECTED TOOLS:** [List of tools you estimate are needed to accomplish the task] **Strategy Evaluation Results (0.00-0.99 scale):** [Show your evaluations like:] - TechniqueName: solution=X.XX, efficiency=Y.YY, total=Z.ZZ ✓ (if ≥1.53) **Selected Cognitive Technique(s):** [List techniques scoring ≥1.53] **Estimated Tools Needed:** [1-8 tools for implementation] --- **Now:** Apply your selected cognitive technique(s) to actually solve the original problem "${input}" using your enhanced reasoning framework.`; }
• src/index.ts:158-170 (schema)

  Input schema definition for the 'deliberate' tool, specifying 'input' as required string and optional 'context' string.
  type: "object", properties: { input: { type: "string", description: "The primary input, question, problem, or task requiring cognitive deliberation", }, context: { type: "string", description: "Optional additional context, background information, or constraints", }, }, required: ["input"], },
• src/index.ts:151-174 (registration)

  Registers the 'deliberate' tool in the MCP server's listTools handler, including name, description, and input schema.
  server.setRequestHandler(ListToolsRequestSchema, async () => { return { tools: [ { name: "deliberate", description: "Advanced cognitive deliberation framework implementing 6-stage processing (Scientific Investigation → OOReD → Critical Thinking → Review → OOReD → Act) with dynamic prompting strategy evaluation. Takes a prompt combining the question/problem and any context, returns comprehensive cognitive processing results with tool usage recommendations.", inputSchema: { type: "object", properties: { input: { type: "string", description: "The primary input, question, problem, or task requiring cognitive deliberation", }, context: { type: "string", description: "Optional additional context, background information, or constraints", }, }, required: ["input"], }, }, ], }; });
• src/index.ts:180-199 (handler)

  Dispatch handler in CallToolRequestSchema that validates arguments and invokes the DeliberationEngine.deliberate method to execute the tool.
  if (name === "deliberate") { const { input, context } = args as { input: string; context?: string }; if (!input || typeof input !== "string") { throw new Error("Input is required and must be a string"); } try { const result = deliberationEngine.deliberate(input, context); return { content: [ { type: "text", text: result, }, ], }; } catch (error) { throw new Error(`Deliberation failed: ${error}`); }
• src/index.ts:11-63 (helper)

  Helper constant defining cognitive prompting strategies used within the deliberate tool's prompt generation.
  const PROMPTING_STRATEGIES = { "Chain of Draft (CoD)": { description: "Concise reasoning drafts ≤5 words/step. Essential calculations only. Abstract verbose details." }, "Cache-Augmented Reasoning + ReAct": { description: "Interleave knowledge activation with reasoning cycles. Keep rationale concise (≤8 bullets). Progressive knowledge building." }, "Self-Consistency": { description: "Generate 3 reasoning drafts in parallel. Return most consistent answer for high-stakes decisions." }, "PAL (Program-Aided Language)": { description: "Generate executable code for computational tasks. Include result + minimal rationale. Prefix '# PoT offload'." }, "Reflexion": { description: "Single critique and revision cycle. Use when confidence < 0.7. Avoid verbose chain-of-thought exposure." }, "Context-Compression": { description: "LLMLingua compression when context exceeds budget. Prefer Minimal-CoT and bounded ToT-lite." }, "ToT-lite (Tree of Thoughts)": { description: "Bounded breadth/depth exploration. Limited branching for complex problem decomposition efficiency." }, "Metacognitive Prompting (MP)": { description: "5-stage introspective reasoning: understand → judge → evaluate → decide → assess confidence. Human-like cognition." }, "Automated Prompt Optimization (APO)": { description: "Autonomously refine prompts via performance feedback. Expert prompting + iterative refinement. Reduces manual effort." }, "Reflexive Analysis": { description: "Embed ethical/legal/cultural considerations. Evaluate against project guidelines. Indigenous Data Sovereignty aware." }, "Progressive-Hint Prompting (PHP)": { description: "Use previous outputs as contextual hints. Multi-turn interaction with cumulative knowledge building." }, "Cache-Augmented Generation (CAG)": { description: "Preload relevant context into working memory. Eliminate real-time retrieval dependencies." }, "Cognitive Scaffolding Prompting": { description: "Structure reasoning through metacognitive frameworks. Mental model construction + validation. Self-monitoring processes." }, "Internal Knowledge Synthesis (IKS)": { description: "Generate hypothetical knowledge constructs from parametric memory. Cross-reference internal knowledge consistency." }, "Multimodal Synthesis": { description: "Process text/images/data integration. Visual question answering + cross-modal analysis. Broader task solutions." }, "Knowledge Synthesis Prompting (KSP)": { description: "Integrate multiple internal domains. Fine-grained coherence validation. Cross-domain knowledge integration." }, "Prompt Compression": { description: "LLMLingua for token budget management. Preserve semantic content while reducing length constraints." } };