ClipSnap defines 2 tools with reasonable descriptions and basic schemas. paste_recent has a strong, context-rich description (180+ chars) explaining when to use it and what it does. cleanup_images description is adequate but generic. Both tools have input schemas with type constraints (min/max for numbers), but lack comprehensive parameter descriptions, most params have minimal guidance. No output schema is documented. Error handling is present but minimal (returns isError flag without recovery guidance). Tool naming follows verb_noun convention (paste_recent, cleanup_images) and is clear. Security annotations (readOnlyHint) are present on paste_recent. Overall: solid foundation but missing parameter-level detail and output documentation expected of production tools.

• Add detailed parameter descriptions: 'folder' should specify 'Absolute or tilde-expanded path (e.g., ~/Pictures). If omitted, scans session history + Spotlight.'; 'max_dimension' should explain 'Resize images so the longest edge does not exceed this pixel count. Reduces token usage for large screenshots.'
• Document the output schema for paste_recent: specify that responses contain alternating image blocks (base64 PNG/JPEG) and text metadata blocks with filename, dimensions, and timestamp.
• Enhance cleanup_images description: 'Manually delete temporary images from the current session. Automatic cleanup runs every 24h or when storage exceeds 500MB. Use all=true only to purge all sessions (irreversible).'
• Implement a dry-run mode for cleanup_images: add a 'preview=true' parameter that returns a list of files to be deleted without actually deleting them, allowing the LLM to confirm before executing.
• Categorize errors with recovery hints: return structured error objects with type (not_found, permission_denied, invalid_input) and next_action (e.g., 'Try search_users() first' or 'Check folder path').