Four tools with basic descriptions and minimal parameter documentation. All tools have descriptions (10-50 chars baseline), but lack depth on WHEN to use them and what they return. Input schemas present but sparse: list_documents has empty input {}, others have 1-2 string params with minimal constraint info. No output schemas documented. Error handling returns permission-denied strings but lacks recovery guidance or actionable next steps. No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Naming is clear (verb_noun pattern) but descriptions are too terse to guide LLM selection effectively.

• Add output schemas to all tools. E.g., list_documents returns {documents: [{id: string, title: string}], total: number}. read_document returns {title: string, content: string}.
• Expand descriptions to 50-150 chars with WHEN/WHY context. E.g., 'Read a document's full content. Call after list_documents() to retrieve text. Requires read permission.'
• Replace example values in parameter descriptions with enum/pattern constraints. E.g., document_id: {type: string, pattern: '^[a-z0-9-]+$', description: 'Document ID (lowercase alphanumeric and hyphens)'}.
• Add tool annotations: readOnlyHint=true for list_documents and read_document; destructiveHint=true for edit_document and approve_refund; idempotentHint=true for read_document.
• Enhance error responses with recovery guidance. E.g., 'Permission denied: user:bob does not have read access to document:q1-report. Call list_documents() to see accessible documents, or contact [email protected] for access.'