Thoughtbox

/** * Interleaved Thinking Resource Templates * * Provides mode-specific IRCoT-style reasoning guides that can be accessed * via resource templates (thoughtbox://interleaved/{mode}) */ /** * Valid modes for interleaved thinking guides */ export type InterleavedMode = "research" | "analysis" | "development"; /** * Abstract capability categories used in mode configurations */ export type CapabilityKind = | "thoughtbox_workspace" | "retrieval_search" | "code_repo" | "sandbox_execute"; /** * Configuration for each mode */ export interface ModeConfig { mode: InterleavedMode; title: string; description: string; requiredCapabilities: CapabilityKind[]; optionalCapabilities: CapabilityKind[]; phases: string[]; notes: string[]; } /** * Mode-specific configurations */ export const MODE_CONFIG: Record<InterleavedMode, ModeConfig> = { research: { mode: "research", title: "Research Mode", description: "For literature review, information gathering, and synthesis tasks", requiredCapabilities: ["thoughtbox_workspace", "retrieval_search"], optionalCapabilities: [], phases: [ "Tooling Inventory - Ground yourself in available search and retrieval capabilities", "Tooling Sufficiency Assessment - Verify you have what you need for the research task", "Strategy in Thoughtbox - Plan your research approach using thoughtbox for structured reasoning", "Interleaved Execution Loop - Alternate between thoughtbox reasoning and information retrieval", "Final Answer - Synthesize findings and deliver comprehensive response" ], notes: [ "Research requires reliable access to external information sources", "Use thoughtbox to track source quality and synthesis progress", "Gate your work at sufficiency assessment - if you lack search capabilities, be honest", "Prioritize breadth in initial passes, depth in later iterations" ] }, analysis: { mode: "analysis", title: "Analysis Mode", description: "For deep examination, pattern recognition, and interpretation tasks", requiredCapabilities: ["thoughtbox_workspace"], optionalCapabilities: ["retrieval_search"], phases: [ "Tooling Inventory - Confirm you have thoughtbox and any supplementary tools needed", "Tooling Sufficiency Assessment - Determine if additional context is required", "Strategy in Thoughtbox - Outline your analytical framework and approach", "Interleaved Execution Loop - Alternate between thoughtbox reasoning and focused investigation", "Final Answer - Present insights, patterns, and conclusions" ], notes: [ "Analysis can often proceed with thoughtbox alone", "External retrieval is optional - use only if context is insufficient", "Focus on identifying patterns, relationships, and implications", "Use thoughtbox to maintain analytical rigor and track assumptions" ] }, development: { mode: "development", title: "Development Mode", description: "For code implementation, debugging, and system building tasks", requiredCapabilities: ["thoughtbox_workspace", "code_repo", "sandbox_execute"], optionalCapabilities: ["retrieval_search"], phases: [ "Tooling Inventory - Verify access to codebase, execution environment, and thoughtbox", "Tooling Sufficiency Assessment - Confirm you can read, write, and test code", "Strategy in Thoughtbox - Design your implementation approach and test strategy", "Interleaved Execution Loop - Alternate between thoughtbox planning, coding, and validation", "Final Answer - Deliver working, tested implementation with documentation" ], notes: [ "Development requires ability to modify code and verify changes", "Use thoughtbox for design decisions and debugging strategy", "Gate your work if you cannot execute or test code - don't proceed blind", "Prioritize incremental validation - test early and often" ] } }; /** * Validates if a mode string is a valid InterleavedMode */ export function isValidMode(mode: string): mode is InterleavedMode { return mode === "research" || mode === "analysis" || mode === "development"; } /** * Generates mode-specific interleaved thinking guide content */ export function interleavedGuide(mode: InterleavedMode): string { const config = MODE_CONFIG[mode]; return `# Thoughtbox Interleaved Thinking - ${config.title} ${config.description} ## Overview This guide implements the IRCoT (Interleaved Retrieval and Chain-of-Thought) pattern, adapted for Thoughtbox as your canonical reasoning workspace. Use this approach when ${getModeUsageGuidance(mode)}. ## Required Capabilities Your environment must provide these capability kinds: ${config.requiredCapabilities.map(cap => `- **${cap}** - ${getCapabilityDescription(cap)}`).join('\n')} ${config.optionalCapabilities.length > 0 ? `## Optional Capabilities These may enhance your work but are not strictly required: ${config.optionalCapabilities.map(cap => `- **${cap}** - ${getCapabilityDescription(cap)}`).join('\n')} ` : ''} ## Process Phases ${config.phases.map((phase, idx) => `### Phase ${idx + 1}: ${phase.split(' - ')[0]} ${phase.split(' - ')[1]} `).join('\n')} ## Execution Pattern \`\`\` WHILE task not complete: 1. Use thoughtbox for structured reasoning step 2. Identify next action based on current understanding 3. Execute action using appropriate tool 4. Observe results 5. Return to thoughtbox to integrate findings 6. Reassess approach and next steps END WHILE \`\`\` ## Mode-Specific Guidance ${config.notes.map(note => `- ${note}`).join('\n')} ## Honesty Requirements **CRITICAL**: Do not proceed if you lack required capabilities. - If you cannot access thoughtbox, this entire approach fails - If you lack required mode-specific capabilities, acknowledge this limitation - Never fabricate tool capabilities or fake execution results - It is better to admit constraints than to produce unreliable work ## Self-Check Procedure Before beginning, verify: 1. [ ] I have access to thoughtbox tool ${config.requiredCapabilities .filter(cap => cap !== "thoughtbox_workspace") .map(cap => `2. [ ] I have access to ${cap} capabilities`) .join('\n')} 3. [ ] I understand the task requirements 4. [ ] I have a strategy for validation If any check fails, communicate the limitation clearly before proceeding. ## References - **Pattern**: IRCoT (Interleaved Retrieval and Chain-of-Thought) - **Paper**: https://arxiv.org/abs/2212.10509 - **Workspace**: Thoughtbox MCP Server - **Mode**: ${config.mode} --- *This guide is generated dynamically via the thoughtbox://interleaved/${mode} resource template.* `; } /** * Helper function to provide mode-specific usage guidance */ function getModeUsageGuidance(mode: InterleavedMode): string { switch (mode) { case "research": return "you need to gather, evaluate, and synthesize information from multiple sources"; case "analysis": return "you need to deeply examine patterns, relationships, or implications in existing information"; case "development": return "you need to design, implement, and validate code or system changes"; } } /** * Helper function to describe capability kinds */ function getCapabilityDescription(capability: CapabilityKind): string { switch (capability) { case "thoughtbox_workspace": return "Structured reasoning workspace for step-by-step thinking"; case "retrieval_search": return "Ability to search and retrieve external information"; case "code_repo": return "Access to read and write code in a repository"; case "sandbox_execute": return "Ability to execute and test code safely"; } }