Server has 19 tools with schemas and descriptions present, but quality is inconsistent. Most tool descriptions are adequate (50-150 chars), but many lack actionable context for LLM selection. Parameter descriptions exist but are often generic. Critical gaps: no error handling guidance, no output schemas documented, no tool annotations (readOnlyHint/destructiveHint), and several tools expose API keys as parameters (security violation). Naming is mostly verb-noun compliant but some tools like 'batch_parse' and 'cost_log' lack clarity on when to use them vs alternatives. No pagination support visible despite batch operations. Average tool quality ~42/100, below the 50-60 baseline for community servers.

• Move api_key parameters to server-side environment variables (GOOGLE_AI_API_KEY). Update tool descriptions to state 'API key configured server-side' instead of accepting it as a parameter.
• Document output schemas for all tools. E.g., batch_notify returns {results: [{url, action, notify_time, error}], summary: {success, error}, quota_warning}. Include field types and when fields are null.
• Add tool annotations: mark cost_reset and preset_delete with destructiveHint=true; mark cost_summary, cost_today, preset_list with readOnlyHint=true. This enables LLM safety reasoning.
• Enhance error messages with recovery guidance. E.g., batch_parse should return 'CSV format error at row 5: missing 'prompt' column. Expected columns: url, prompt, model, aspect_ratio' instead of a generic parse error.
• Clarify tool relationships in descriptions. E.g., 'cost_log: Record a single generation cost. Use cost_estimate to forecast batch costs before running; use cost_summary to review total spend.' This prevents LLM confusion between similar tools.