Single tool 'hashid_scan' has a basic description but critically lacks parameter descriptions and output schema documentation. The tool name is acceptable (verb_noun form) but the input schema is minimal, only 'target' with a generic string type and no format constraints. No output schema is documented, forcing LLMs to guess the response structure. Error handling is absent. The implementation is visible in hashid/mcp_server.py but the tool definition is inferred from the @app.tool() decorator rather than explicitly registered with full metadata.

• Add explicit output schema documentation: describe the JSON structure returned by to_json(scan(target)). Include field names, types, and what each field represents (e.g., 'findings' array with 'id', 'severity', 'confidence' fields).
• Add parameter description for 'target': 'Hash string to identify (e.g., "5d41402abc4b2a76b9719d911017c592"). Supports MD5, SHA-1, SHA-256, bcrypt, and other common formats. Length: 1-512 characters.'
• Document expected output fields and add examples in the tool description: 'Returns JSON with fields: hash_type (string), confidence (0-1), crack_feasibility (string: easy|medium|hard|infeasible), estimated_time (string).'
• Add error handling guidance: 'If hash type cannot be identified, returns confidence=0 and suggests similar types. If input exceeds 512 chars, returns error with guidance to split input.'
• Explicitly register the tool with FastMCP using full schema metadata instead of relying on decorator inference. Include input schema with type, description, and optional enum/pattern constraints.