Server has 17 tools with consistent naming (verb_noun pattern) and basic descriptions. All tools have input schemas with type definitions. However, descriptions are terse (avg ~50 chars, well below the 194-char baseline), parameter descriptions lack detail on constraints/formats, and output schemas are not documented. Error handling guidance is absent. Tool composition is sound (single responsibility), but descriptions do not explain WHEN to use each tool or what to expect in responses. Security-sensitive tools (delete_*) lack confirmation patterns.

• Expand tool descriptions to 100 - 200 chars. Include WHAT the tool does, WHEN to use it (vs similar tools), and any prerequisites. E.g., 'Search for links by title, URL, or tags. Returns paginated results with link_id, title, and collection_id. Use get_link_content to retrieve archived text. Requires at least one search term.'
• Add parameter descriptions for all inputs. Specify format, range, and constraints. E.g., 'cursor: Opaque pagination token from previous response. Omit for first page.' 'max_chars: Maximum characters to return (1 - 1000000, default 100000).'
• Document output schemas for all tools. List response fields, types, and which IDs are available for chaining. E.g., search_links returns [{link_id, title, url, collection_id, tags: []}].
• Implement confirmation pattern for destructive tools. Add a 'confirm_delete_link' tool or a 'dry_run' parameter to delete_* tools. Return 'This will permanently delete X. Call confirm_delete_link(token) to proceed.' Agents can then ask the user before executing.
• Add error recovery guidance. E.g., 'Link not found. Try search_links() to find the correct link_id.' 'Invalid collection_id. Call get_collections() to list available collections.' This enables agents to self-correct without dead-ending.