Loading ToolBench
MCP Server for partitioning text into sentence-aware sliding window chunks without breaking grammatical boundaries
Single tool with minimal schema and description. Tool name 'chunk_document' lacks action verb clarity (should be 'chunk_text' or 'partition_text'). Description is 67 chars, acceptable but generic. Input schema present but lacks parameter descriptions entirely. No output schema documented. No error handling guidance. No parameter constraints (max_words is unbounded). Fails multiple critical checks from pattern:tool and pattern:tool-description.
Partition text into sentence-aware sliding window chunks.
Missing parameter descriptions in input schema. 'text' and 'max_words' have no descriptions; LLMs cannot infer what max_words controls or its valid range.
No output schema documented. Tool returns JSON array of chunk objects with fields (chunk_index, word_count, content, chunk_hash), but this structure is not declared in the tool definition. LLMs cannot plan downstream operations.
Unbounded numeric parameter. 'max_words' has no minimum/maximum constraints. LLMs could pass 0, negative, or absurdly large values (e.g., 1000000), causing performance issues or unexpected behavior.
Tool name lacks action verb. 'chunk_document' is ambiguous, does it chunk, split, partition, or segment? Verb-first naming (e.g., 'partition_text', 'chunk_text') helps LLMs infer intent from the name alone.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
No error handling guidance. If chunking fails (empty text, invalid max_words), the tool returns no recovery hints. LLMs cannot self-correct or know what to try next.