The server defines 6 tools with reasonable naming conventions (verb-first, mostly clear intent). All tools have descriptions and input schemas are partially present. However, there are significant gaps: parameter descriptions are minimal or missing, output schemas are not documented, error handling is absent from tool definitions, and no guidance on recovery paths. Tool names follow verb_noun convention (get_*, create_*) which is positive. The create_ai_profile tool has the most detailed schema with nested phase objects, but even this lacks comprehensive documentation of output structure and error cases. Descriptions are adequate (ranging 30-90 chars) but lack clarity on when to use each tool vs. similar ones, what fields are returned, and what errors might occur.

• Document the output schema for each tool. Specify what fields are returned, their types, and when to use them. E.g., get_profiles should return {profiles: [{id: string, name: string, ...}], total: number}.
• Add detailed parameter descriptions with constraints. E.g., 'limit' should be 'Maximum number of shots to return. Must be 1 - 1000, defaults to 50.'
• Define error conditions and recovery paths for each tool. E.g., 'If profile_id is not found, returns error code PROFILE_NOT_FOUND with message suggesting search_profiles() or list of available IDs.'
• Declare idempotency for create_ai_profile. State clearly whether calling it twice with identical parameters returns the same profile or creates a new one.
• Add pagination documentation to get_profiles and get_shot_history. Clarify offset vs. cursor behavior, max result limits, and how to fetch the next page.
• Specify field constraints: is profile_id a UUID, slug, or numeric? Is shot_id URL-safe? Document format expectations so LLMs can validate before calling.
• Add tool annotations (readOnlyHint, destructiveHint) to the tool definitions. get_profiles, get_profile, get_shot_notes, get_shot_history, get_shot_details should be marked readOnlyHint=true; create_ai_profile should be marked destructiveHint=true (modifies state).
• Clarify the relationship between get_shot_notes and get_shot_details. When should the LLM use each? Do they return different fields or overlapping data?