28 tools with inconsistent quality. Most have basic descriptions (50-150 chars) and verb-prefixed names, but critical gaps: NO input schemas visible in source code for ANY tool, all schema scores are 0. Parameters documented in rubric (phone, text, etc.) but not in actual code definitions. Error handling absent. No output schemas documented. No parameter validation guidance. No enum constraints. Descriptions lack WHEN/WHY context. No idempotency markers. Security concerns: phone numbers and message content exposed in logs without sanitization despite PrivacyGuard class.

• Add explicit JSON Schema to every tool via FastMCP @mcp.tool() decorator with input_schema parameter. Include type, description, and constraints (enum, minLength, pattern) for each parameter.
• Expand descriptions to 100-200 chars following pattern: '[WHAT] [WHEN/WHY] [RETURN/SIDE-EFFECT]'. E.g. 'Send a text message to a WhatsApp contact. Use this to initiate conversations or reply to existing chats. Returns message_id and timestamp.'
• Document output schemas inline or in docstrings. E.g. whatsapp_get_messages should specify: returns array of {message_id, sender, text, timestamp, media_type?}.
• Add error handling guidance to descriptions. E.g. whatsapp_send_text: 'If phone is invalid, returns error with suggestion to call whatsapp_search_contacts first. If rate-limited, retry after 60s.'
• Fix PrivacyGuard.log_action() to redact 'phone' and 'message_id' params in addition to message content. Extend PATTERNS to include phone number format and apply redaction to all sensitive param keys.

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

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