Loading ToolBench

Reading the current corpus…

ai-assistant MCP server · F · ToolBench

ai-assistant

An AI assistant that manages emails and calendar events using Google APIs, with support for multiple LLM backends (Gemini, Ollama, OpenAI) and both interactive (Gradio) and non-interactive (ADK) interfaces.

<=2025-11-25STDIO17 tools

47/100

Definition Quality50

Protocol Readiness47

Supportability43

Spec posture

Inferred effective spec <=2025-11-25·current baseline 2026-07-28

Static source inference · medium confidence · detected: Logging, Dynamic Client Registration

Deprecated protocol patterns detected

Summary

Server has 17 tools with basic schemas and descriptions, but significant gaps limit production readiness. Most tools have descriptions (10-150 chars, below the 194-char baseline), and input schemas are present but often lack parameter descriptions. Critical issues: Ask User tool has no actionable description; many parameters lack type constraints (enums, ranges); no output schemas documented; error handling is absent; no security/permission gates; composition issues (list_gmail_tools and list_calendar_tools are discovery-only, not actionable). STDIO transport caps protocol score at 50. Average per-tool score: 52.

Tools (17)

Ask Userread onlysource verified50/100

Ask the user a question when you need clarification or additional information. Input: question string.

create_event_toolwriteauth50/100

Create a calendar event with ISO8601 start/end times.

delete_event_tooldestructiveauth50/100

Delete a calendar event by event_id.

find_meeting_slots_toolread onlyauth50/100

Find available meeting slots for multiple attendees.

get_emails_toolread onlyauthsource verified70/100

Search Gmail by query, with pagination and optional full body.

get_events_toolread onlyauth50/100

Fetch Google Calendar events with optional filters.

get_free_busy_toolread onlyauth50/100

Get free/busy blocks for one or more calendars in a time range.

get_thread_tool

Issues (5)

HIGHtool-description ↗

Ask User tool description is vague ('Ask the user a question when you need clarification or additional information. Input: question string.'). Does not explain when to use it, what happens with the response, or how it integrates with the workflow. LLM cannot determine if this is for confirmation, data collection, or error recovery.

Ask User

CRITICALresponse-shaper ↗

No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields. E.g., get_emails_tool returns emails but schema does not specify fields (id, subject, from, snippet, etc.), forcing LLM to guess structure.

get_emails_toolget_thread_toolget_events_toolfind_meeting_slots_toolget_free_busy_toollist_calendars_tool

HIGHconstrained-input ↗

Many parameters lack type constraints. E.g., send_email_tool 'to' and 'cc' accept strings but no validation for email format; create_event_tool 'recurrence' is array but no documentation of RRULE format; get_emails_tool 'gmail_query' accepts free-form string with no guidance on valid Gmail search syntax.

Recommendations

Add output schemas to all tools. Document the structure of returned objects (e.g., email has id, subject, from, snippet, body; event has id, summary, start, end, attendees). This enables LLM to chain tools and extract required fields.
Enhance parameter descriptions with format constraints. E.g., 'to' parameter: 'Recipient email address(es), comma-separated (e.g., [email protected] or [email protected],[email protected])'. 'recurrence' parameter: 'Array of RRULE strings (e.g., ["RRULE:FREQ=WEEKLY;COUNT=10"])'. 'gmail_query': 'Gmail search query syntax (e.g., "from:[email protected] is:unread").'
Add error handling and recovery guidance to write operations. E.g., send_email_tool: 'Returns error if recipient email is invalid or account lacks send permission. If draft_mode=true, error is unlikely. If draft_mode=false and send fails, try draft_mode=true first.'
Implement confirmation pattern for destructive operations. delete_event_tool should accept a 'confirm' boolean parameter (default false) or return a confirmation token requiring a second call. Prevents accidental deletion.
Improve Ask User tool description. E.g., 'Ask the user for clarification or additional information when the current context is ambiguous. Returns the user's response as a string. Use when you need confirmation before taking action (e.g., "Which email thread?") or when required parameters are missing.'
Add permission/scope declarations to tools. E.g., send_email_tool requires 'write:email' scope; delete_event_tool requires 'write:calendar' and 'delete:calendar'. Enables least-privilege agent configuration.

Spec posture evidence

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

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

Score history

Overall score trend

First recorded score · v2 rubric

47/100

Scored	Grade	Overall	Spec posture	Rubric
2026-09-23	F	47	<=2025-11-25	v2