Loading ToolBench
Server has 6 tools with complete input schemas and descriptions present. Naming follows verb_noun convention (calendar_today, gmail_list_unread, gmail_send). However, descriptions are minimal (10-50 chars), lacking context on WHEN to use each tool, dependencies, and return value structure. Parameters are typed but lack detailed constraints (e.g., max_results accepts 1-100 but description doesn't state bounds). No output schemas documented. Error handling and recovery guidance absent. Account parameter is optional but behavior when omitted is not explained. Tools lack idempotency hints despite write operations (calendar_create_event, gmail_send).
Create a calendar event.
Get today's calendar events.
Get this week's calendar events (Monday through Sunday).
Get full email content by message ID.
List unread emails with subject, sender, date, and snippet.
Send an email.
Descriptions are too brief (10-50 chars). Missing context on WHEN to use each tool, what it returns, and dependencies. LLMs cannot determine tool selection without richer descriptions.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., event_id from calendar_create_event, message_id from gmail_send).
Write operations (calendar_create_event, gmail_send) lack idempotency hints and confirmation/dry-run support. Agents cannot safely retry without risking duplicate events or emails.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 53 | 2026-07-28+ | v2 |
Parameter constraints not documented in descriptions. max_results accepts 1-100 but description omits bounds. ISO 8601 format required for start/end times but no validation guidance provided.
Optional 'account' parameter behavior undefined. What happens if omitted? Does it use first account, fail, or require env config? LLMs need explicit guidance.