All 14 tools have basic descriptions (10-50 chars) and input schemas with type definitions. However, descriptions are uniformly terse and lack LLM-optimized guidance on WHEN to use each tool, WHAT it returns, or HOW it chains with others. Parameter descriptions are minimal (e.g., 'Session ID (optional, defaults to current session)' lacks context on what a session represents). No output schemas are documented, LLMs cannot infer what fields to expect from responses. All tools are READ_ONLY, which is good for safety, but error handling and recovery guidance are absent. Tool naming follows verb_noun convention (get_*, list_, suggest_*), which is correct, but lacks distinction between similar tools (e.g., get_session_summary vs get_project_stats both aggregate data but serve different scopes).

• Document output schemas for all tools. For list_sessions, specify: returns array of {session_id, timestamp, token_count, cache_hit_rate, status}. For get_budget_status, specify: returns {daily: {used, limit, remaining}, weekly: {...}, monthly: {...}}. This enables LLMs to chain tools and extract the right fields.
• Expand tool descriptions to 100-200 chars with WHEN/WHAT/HOW guidance. E.g., get_session_summary: 'Retrieve aggregated token usage, cache efficiency, and cost metrics for a session. Use this to understand overall session performance before diving into detailed history. Returns summary stats and links to related tools (get_history, get_cache_efficiency).'
• Add context to parameter descriptions. E.g., session_id: 'Unique identifier for a session (UUID format). Defaults to the current active session. Obtain from list_sessions() if you need to query historical data.' This helps LLMs decide when to provide vs omit optional params.
• Add pagination guidance to list_sessions and get_interaction_history. Specify: 'Returns up to {limit} items. If more exist, include a next_cursor or total_count in response so agents can fetch additional pages without guessing.'
• Clarify tool relationships in descriptions. E.g., get_session_summary: 'High-level overview. For detailed token breakdown, call get_history(). For cache metrics, call get_cache_efficiency(). For cost analysis, call get_budget_status().' This guides multi-step planning.

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

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