Loading ToolBench
A server to create drafts of X(twitter) posts, threads using LLMs and post directly from the chat.
Server has 5 tools with explicit schemas and basic descriptions, but falls short of production quality. All tools have input schemas (positive), but descriptions are minimal (10-30 chars, well below the 50-200 char LLM-optimized baseline). Parameter descriptions are present but generic. No output schemas documented. Error handling is minimal, tools raise generic RuntimeError/ValueError without recovery guidance. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite having WRITE, IRREVERSIBLE, and DESTRUCTIVE risk levels. No pagination support for list_drafts. Missing dependency hints and state-change clarity in descriptions.
Create a draft tweet thread
Create a draft tweet
Delete a draft tweet or thread
List all draft tweets and threads
Publish a draft tweet or thread
Descriptions are too short (10-30 chars) to guide LLM tool selection. Baseline is 50-200 chars. Missing WHEN to use, WHAT it returns, and state-change implications.
No tool annotations despite explicit risk levels (WRITE, IRREVERSIBLE, DESTRUCTIVE). publish_draft and delete_draft need destructiveHint=true; create_* need idempotentHint clarification.
No output schemas documented. LLMs cannot plan downstream calls or extract required fields (e.g., draft_id from create_draft_tweet response).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
Error handling is generic (RuntimeError, ValueError) with no recovery guidance. publish_draft and delete_draft lack confirmation/dry-run pattern for irreversible operations.
list_drafts has no pagination (limit, offset, cursor) or result count. Unbounded results risk context window exhaustion.