Two tools with basic schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Tool names follow verb_noun convention (perplexity-query, perplexity-search), which is good. However, descriptions are minimal (34-45 chars, below the 50-200 char baseline for LLM optimization). Input schemas use Zod with enums for model selection, but lack detailed parameter descriptions beyond single-line text. No documented output schema structure. Error handling returns generic text responses without recovery guidance. The tools are read-only (low risk), but lack the depth expected for production use.

• Expand tool descriptions to 80-150 chars. Example: 'perplexity-query: Send a direct question to Perplexity for reasoning-based answers. Use when you need deep analysis or multi-step reasoning. Returns text response with citations.' vs. current 34 chars.
• Document output schema explicitly. Add to tool docstring: 'Returns: {content: [{type: "text", text: string}]}. The text field contains the model response with inline citations.'
• Add parameter descriptions with constraints. Example for systemPrompt: 'Optional system prompt (max 500 chars) to set context or role. E.g., "You are a research assistant." Omit for default behavior.'
• Implement error classification in handleError(). Return {content: [{type: 'text', text: string}], isRetryable: boolean, errorType: 'api_error'|'auth_error'|'rate_limit'} to guide LLM recovery.
• Clarify tool differentiation in descriptions. perplexity-query: 'For direct Q&A and reasoning.' perplexity-search: 'For current web information and fact-checking with citations.'

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

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