Server has 5 tools with basic descriptions and some parameter documentation, but significant gaps in schema completeness, parameter descriptions, and error handling. Tool names follow verb_noun convention (chat_with_gemini, get_model_info, get_health_status), which is good. However, parameter descriptions are sparse or missing entirely. The chat_with_gemini tool lacks output schema documentation. No error handling guidance is present, tools will fail silently or return generic exceptions. Descriptions are present but brief (avg ~60 chars), below the 194-char baseline for A+ tools. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite all being read-only operations.

• Add explicit output schemas for all tools. E.g., get_model_info should document: {model: string, project: string, region: string, preferred_context_size: int, max_context_size: int, rate_limit: string}. chat_with_gemini should document: {content: string, finish_reason: string, usage?: {input_tokens: int, output_tokens: int}}.
• Expand parameter descriptions with format, range, and constraints. E.g., 'temperature: float between 0.0 (deterministic) and 1.0 (random). Default 0.7 balances coherence and creativity.' 'max_tokens: integer, 1 - 32000. If omitted, uses model default.'
• Add error handling guidance. E.g., 'If authentication fails, verify GCP_PROJECT_ID and credentials. If rate-limited, retry after 60 seconds. If model not found, call get_model_info to list available models.'
• Annotate all tools with readOnlyHint=true since they are all read-only queries. This enables LLM optimizations and signals safety for caching.
• Expand tool descriptions to 100 - 200 chars explaining WHEN to call each. E.g., 'Get current model info to check available context size before sending large prompts. Useful for planning multi-turn conversations or validating model capabilities.'

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry