Loading ToolBench
AI agent and MCP server layer of the AtlasMind platform: clarification agent and executive briefing engine
Four tools with reasonable descriptions (100-200 chars) and basic parameter schemas. All tools are READ_ONLY and well-intentioned for a Jira/briefing domain. However, parameter descriptions are sparse or missing entirely (e.g., 'limit' in query_jira lacks guidance on valid range; 'session_id' across tools lacks explanation of lifecycle). Output schemas are not documented, LLMs cannot predict what fields to expect from responses. Error handling is mentioned but not detailed in tool definitions. No tool annotations (readOnlyHint, idempotentHint) despite all being read-only. Naming is clear (verb_noun pattern), but composition could be tighter, clarify() creates a stateful session dependency that complicates agent planning.
Continue a clarification session. Accepts a user response to a clarification question and returns the next question or the final query result.
Generate an executive briefing from an agenda. Decomposes topics, queries Jira for issues per topic, analyzes issues, ranks them, and synthesizes a markdown report.
Retrieve a previously generated report by ID. Returns the markdown content and metadata.
Query Jira issues with clarification support. Accepts a natural-language query, optionally clarifies ambiguous terms through multi-round elicitation, and returns ranked issues with analysis.
Output schemas not documented. LLMs cannot predict response structure (fields, types, pagination). Responses likely include verbose metadata (Jira raw fields, internal IDs) that should be stripped.
Parameter descriptions incomplete. 'limit' lacks range guidance (e.g., 1 - 100); 'session_id' lacks lifecycle explanation (generated if omitted, but how long valid?); 'response' in clarify() is vague about expected format.
Stateful session management (session_id) complicates agent planning. Multi-round clarification requires agents to track session state across calls. Consider wrapping into single tool or documenting session lifecycle explicitly.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | <=2025-11-25 | v2 |
No tool annotations despite all tools being read-only. Add readOnlyHint=true to all four tools to signal safety to agents and prevent unnecessary confirmation prompts.
Error handling not visible in tool definitions. No guidance on what errors are retryable, what user input is needed, or how to recover from ambiguous queries.