Dataset Genie exposes 6 tools via fastmcp/HTTP with minimal documentation. Tool names follow verb_noun convention (health, secrets_status, set_secret, get_settings, update_settings, list_models), which is good. However, descriptions are extremely brief (10-50 chars), parameter descriptions are absent or minimal, and output schemas are not documented. No error handling guidance, no parameter constraints (enums, ranges), and no indication of which operations are destructive. The set_secret and update_settings tools modify state but lack confirmation patterns or clear side-effect documentation. Input schemas exist but are sparse, most parameters lack descriptions. This is a functional but poorly documented tool set that would confuse LLMs about when to use each tool and what to expect.

• Expand tool descriptions to 100-200 chars. Include WHAT the tool does, WHEN to call it, and what it returns. E.g., 'health: Check if the Genie server is running and retrieve its version. Call this first to verify connectivity before other operations. Returns {ok: bool, version: string, mcp: bool}.'
• Add descriptions to all parameters. For set_secret(name, value): 'name: Secret identifier (e.g., OPENROUTER_API_KEY). Must match a known secret key.' 'value: The secret value to store. Will be encrypted and persisted.'
• Document output schemas inline or in a separate schema section. E.g., 'list_models returns {models: [{id, name, pricing, context_length, ...}], warning: string|null}. Limit is 50 models per call.'
• Add error handling guidance. E.g., set_secret: 'If the secret name is invalid, returns {error: "bad_request", message: "Unknown secret: {name}"}. Valid secrets: OPENROUTER_API_KEY, HF_TOKEN, etc.'
• Mark destructive operations with clear language. E.g., update_settings: 'WARNING: This modifies application settings and may affect running operations. Consider calling get_settings first to review current values.'