DollhouseMCP

# DollhouseMCP Capability Index ALWAYS check this index FIRST. It provides 97% token reduction. ## Layer 1: Element Search Hierarchy ELEMENT_SEARCH_HIERARCHY: DEFAULT ORDER (when location unspecified): 1. Active (already loaded) - 0 tokens 2. Local (~/.dollhouse/portfolio) - 50 tokens 3. GitHub (user's portfolio) - 100 tokens 4. Collection (community library) - 150 tokens OVERRIDE: User intent always takes precedence - "search the collection for..." → Go directly to collection - "check my GitHub for..." → Go directly to GitHub portfolio - "look in my local..." → Go directly to local portfolio - "is there an active..." → Check only active elements ## Layer 2: Tool Capabilities TOOL_CAPABILITIES: search_portfolio: FINDS elements in local storage search_collection: FINDS elements in community library portfolio_element_manager: MANAGES GitHub portfolio sync get_active_elements: CHECKS what's currently loaded activate_element: LOADS element into context create_element: CREATES new element edit_element: MODIFIES existing element list_elements: LISTS available elements by type validate_element: VERIFIES element correctness ## Layer 3: Element Capabilities ELEMENT_CAPABILITIES: memories: PROVIDE: Contextual information on topics PERSIST: Information across sessions AUGMENT: Current context with history personas: ALTER: Behavioral patterns PROVIDE: Specialized expertise SHAPE: Response style skills: PROVIDE: Specific capabilities EXECUTE: Defined procedures ENHANCE: Task performance ## Layer 4: Workflow Processes WORKFLOW_PROCESSES: "I need information about X" → FIRST: Check active memories THEN: Use search_portfolio THEN: Use portfolio_element_manager FINALLY: Use search_collection "Help me debug" → CHECK: Active skills/personas SEARCH: Local portfolio SEARCH: GitHub portfolio SEARCH: Collection "Remember this" → CHECK: Active memories IF_EXISTS: Use edit_element IF_NOT: Use create_element ENSURE: activate_element CRITICAL: This index structure provides 97% token reduction compared to loading all elements.