25 tools with consistent verb-noun naming (list_, get_, create_, set_). All have descriptions (avg 45 chars, baseline 194). Input schemas present with typed parameters. However, descriptions are terse and lack LLM-optimized guidance on WHEN to use each tool or what happens next. Output schemas undocumented. Write operations lack confirmation/dry-run patterns. No error recovery guidance. Parameter descriptions are minimal (avg 20 chars, baseline 72). Missing pagination details for list operations. confirmToken pattern suggests confirmation flow exists but is not formally documented in tool definitions.

• Document output schemas for all tools. For each tool, add a 'Returns' section specifying field names, types, and whether fields are always present. Example: 'Returns: {course_id: number, title: string, enrollment_status: "active"|"inactive", instructor_name: string}'.
• Expand tool descriptions to 100-200 characters. Include WHEN to use the tool, any prerequisites (e.g., 'Requires instructor role'), and what the response enables. Example: 'List all assignments in a course. Returns active and draft assignments if you have instructor role; students see only published assignments. Use this to discover assignment IDs before calling get_assignment.'
• Add 50-100 character descriptions to every parameter. Specify format, range, and how to obtain the value. Example: 'course: Course ID (numeric, 1-999999). Obtain from list_courses() or the course URL.'
• Add pagination to list_* tools: limit (1-100, default 20), offset (0+, default 0), and return total_count and has_more. Document in descriptions: 'Returns up to 20 items per page. Use offset to fetch subsequent pages.'
• Document confirmToken flow in tool descriptions. Specify: 'This tool requires a confirmation token from a preview step. If token is invalid or expired, returns 'Confirmation failed: token expired. Call preview_* to generate a new token.' Implement preview_* tools that return tokens without side effects.