A real-time team chat application with MCP tools for AI assistant integration. Exposes chat operations (post message, search, add reaction) as MCP tools.
WorkChat MCP has 3 tools with basic schemas and descriptions, but significant gaps limit production readiness. All tools have descriptions (10-50 chars, below the 50-200 char baseline for LLM optimization) and input schemas with types. However, descriptions lack context on WHEN to use each tool, what they return, and error recovery paths. Parameter descriptions are minimal (e.g., 'Email of user posting' lacks format guidance). No output schemas documented. Error handling is present in business logic but not surfaced in tool definitions. Security concern: user_email defaults to '[email protected]' across all tools, creating a privilege escalation risk if the default is used unintentionally. Tool composition is reasonable (each does one thing), but naming could be more specific (e.g., 'post_message' vs 'send_message_to_channel').
Add a reaction to a message.
Post a message to a channel.
Search messages using full-text search.
Descriptions are too brief (10-50 chars) and lack LLM-optimized context. 'Post a message to a channel' does not explain WHEN to use this vs other communication tools, what the return value is, or what errors might occur. Baseline for A+ tools is 50-200 chars with WHAT/WHEN/RETURN guidance.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., does search return message IDs, channel IDs, timestamps?). post_message returns a string (message ID), but search and add_reaction return types are not specified.
user_email parameter defaults to '[email protected]' on all three tools. This is a privilege escalation risk: if an agent omits the parameter, all actions execute as admin. Defaults must not cause unintended side effects. Should either require the parameter or default to the authenticated user context.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 45 | 2026-07-28+ | v2 |
Parameter descriptions lack format/constraint guidance. 'Email of user posting' does not specify format (e.g., RFC 5322 email), length limits, or validation rules. 'Emoji reaction' does not clarify if multi-character sequences are allowed or if only single Unicode codepoints are valid.
Error handling is implemented in business logic (e.g., 'Message body cannot be empty', 'Invalid channel ID format') but not surfaced in tool definitions. LLMs cannot see these constraints and will not know what errors to expect or how to recover. Error responses must guide the LLM to the next action.