Severe quality gaps across all 43 tools. All tools lack input parameter descriptions, the rubric requires every parameter to have a description explaining what it controls. Query parameters are documented only as generic 'Query parameters for filtering X' with no detail on valid fields, formats, or constraints. Data parameters are bare dicts with no schema. No output schemas documented. Descriptions are minimal (10-30 chars) and do not explain WHEN to use each tool or what it returns. No error handling guidance. No tool annotations (readOnlyHint, destructiveHint). Tools are directly exposed API wrappers with no agent-friendly abstraction.

• Add detailed descriptions to every parameter. For query_params, document valid filter field names (e.g., 'Filter by status, priority, created_date. Format: status=open&priority=high'). For data params, define required/optional fields and their types.
• Document output schemas for all tools. Specify which fields are returned, their types, and which IDs/references are needed for chaining (e.g., 'Returns: {id, name, program_id, status, created_at}').
• Expand tool descriptions to 50-200 characters. Include WHAT the tool does, WHEN to use it (vs similar tools), and what it returns. For write/delete tools, explicitly state 'This creates/modifies/deletes a resource.'
• Add tool annotations: mark all GET tools with readOnlyHint=true, all DELETE tools with destructiveHint=true, and idempotent tools with idempotentHint=true. This prevents LLMs from retrying destructive operations.
• Implement input validation and error recovery guidance. Return actionable errors like 'Invalid status: got "pending", must be one of: open, in_progress, resolved, closed' instead of raw HTTP errors.