32 tools across 5 integrations (Ariba, Coupa, Jira, Salesforce, SuccessFactors). Tool naming follows verb_noun convention (get_, list_, create_, update_, approve_, reject_). Descriptions present but generic (avg ~60 chars, well below 194-char baseline). Input schemas visible with type definitions and parameter descriptions. Critical gaps: no output schemas documented, no error handling guidance, no pagination details despite list_ tools, no tool annotations (readOnlyHint/destructiveHint), duplicate tool names across modules (tool_list_invoices appears in both Ariba and Coupa), no idempotency guarantees for write operations. Mock data fallback in Ariba suggests incomplete API integration. Descriptions lack WHEN/WHY context and recovery guidance.

• Namespace duplicate tool names by system: rename tool_list_invoices to ariba_list_invoices, coupa_list_invoices, etc. Or use a single unified tool that accepts a 'system' parameter (ariba|coupa|salesforce).
• Document output schemas for all tools. Specify which fields are returned, their types, and which are required for chaining (e.g., invoice_id from list_invoices must be in response to enable reject_invoice calls).
• Add tool annotations: mark all READ_ONLY tools with readOnlyHint=true; mark WRITE tools (reject_invoice, approve_requisition, add_comment, transition_issue, create_task, update_task) with destructiveHint=true and idempotentHint=false unless confirmed idempotent.
• Expand descriptions to 100-200 chars with WHEN/WHY context. Example: 'Get the current payment status of an invoice in SAP Ariba. Use this to check if an invoice has been paid or is pending approval. Returns status, amount, and due date.'
• Add pagination to list tools: include limit (default 20, max 100) and offset/cursor parameters. Document max results in description. Return total_count and next_cursor in response to enable multi-page queries.

Inferred effective spec: <=2025-11-25.

• Relies on Logging (deprecated) - log to stderr or use OpenTelemetry