Server has 9 tools with basic descriptions and input schemas, but significant gaps limit production readiness. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization), and input schemas are present but minimal. Parameter descriptions are present but terse. No output schemas documented. Error handling is absent, tools return no guidance on failure modes or recovery. Destructive tools (set_rating, set_label, set_caption, set_title, set_pick_flag, add_keywords, remove_keywords) lack confirmation/dry-run patterns. Tool names follow verb_noun convention (good), but descriptions do not explain WHEN to use each tool or what it returns.

• Expand all tool descriptions to 50 - 200 characters. Include WHAT the tool does, WHEN to use it (e.g., 'Call after selecting photos'), and WHAT it returns. Example: 'Set the star rating (0 - 5) for currently selected photos. Returns success status and count of photos updated. Requires at least one photo selected.'
• Document output schemas for all tools. Specify the structure of responses (e.g., get_studio_info returns {catalog_name: string, catalog_path: string, plugin_version: string}). This enables LLMs to chain tools and extract the right fields.
• Add confirmation/dry-run pattern to destructive tools. Implement a 'dry_run' boolean parameter (default false) or a separate 'preview_*' tool that shows what would change before committing. Example: set_rating with dry_run=true returns 'Would update 5 photos to 4 stars' without modifying.
• Enhance parameter descriptions with constraints and examples. Instead of 'Integer between 0 and 5', write 'Star rating: 0 (unrated) to 5 (best). Default: 0. Example: 3 for three-star rating.' For enums, use JSON Schema enum constraint instead of prose.
• Add error handling and recovery guidance. Return structured errors like {error: 'NO_SELECTION', message: 'No photos selected. Call get_selection() first to verify photos are selected.', retryable: false}. This guides LLM recovery.

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

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