All 5 tools lack input schemas entirely, cannot verify parameter types, constraints, or structure. Descriptions are present but generic (10-50 chars), providing minimal LLM guidance on when/why to call each tool. No output schemas documented. Tool names use 'mingcang_' prefix but lack action verbs (get_, read_, list_) that signal intent. No error handling, recovery guidance, or parameter validation visible. Composition is unclear, tools appear to read overlapping context (project_context, memory_context, stock_context) without clear differentiation.

• Add explicit JSON Schema input definitions for all tools. Include type, required, description, and constraints (enum, pattern, min/max) for every parameter. Example: {"symbol": {"type": "string", "description": "Stock ticker symbol (e.g., AAPL, MSFT). Required to fetch context.", "required": true}}
• Rename tools with action verbs: read_project_context, get_memory_snapshot, fetch_memory_context, fetch_stock_context, check_health. This signals intent to LLMs before they read descriptions.
• Expand descriptions to 100-200 chars. Explain WHAT the tool returns, WHEN to call it (vs similar tools), and any prerequisites. Example: 'Fetch current project startup context including memory counts, watchlist, and open positions. Call this first to understand the agent state before analyzing specific stocks.'
• Document output schemas for each tool. Specify returned fields, types, and which fields are safe to pass to downstream tools. Example: mingcang_project_context returns {memory_count: int, watchlist: [string], positions: [{symbol: string, quantity: int, ...}]}
• Clarify tool composition. Add a discovery tool (list_available_contexts or describe_context_types) that explains when to call project_context vs memory_context vs stock_context, or merge overlapping tools into one parameterized tool.