Loading ToolBench
A Model Context Protocol server that fetches random images from moe.jitsu.top API.
Single tool with complete input schema and detailed description. Naming is clear (get_random_pic). Description is comprehensive (194 chars) and explains the tool's purpose, parameters, and caveats. Input schema is well-structured with proper types, enums, and defaults. However, output schema is incomplete, it references definitions that are not fully specified in the actual tool registration, and the schema appears to be documentation rather than a machine-parseable JSON Schema. No error handling guidance or recovery patterns documented. Parameters lack individual descriptions in some cases (e.g., 'sort' enum values are not explained). Missing output field documentation for what the response actually contains.
Fetches random images from the moe.jitsu.top API. Specify category, size, and format. By default, returns the image URL in Markdown format (e.g., ''). If you need the actual image data (e.g., for analysis), set the 'include_base64' parameter to true. This will return the image data directly as Base64 encoded content but significantly increases token usage. Note: Returned image URLs might contain escaped slashes (\/\/) due to anti-scraping measures; the client should replace them with a single slash (/).
Output schema is incomplete and not machine-parseable. The outputSchema references definitions that are not fully integrated into the tool registration. Actual response structure (TextContent vs ImageContent union) is documented in comments but not in a formal, parseable schema.
No error handling guidance. Tool description does not explain what happens on API failures, rate limits, or invalid categories. LLM has no recovery path if the API is unavailable or returns unexpected data.
Enum values in 'sort' parameter lack descriptions. The categories ['all', 'mp', 'pc', 'silver', 'furry', 'r18', 'pixiv', 'jitsu'] are not explained, LLM cannot infer what 'mp' or 'pc' mean without external knowledge.
Base64 encoding feature significantly increases token usage but no guidance on when to use it. Description warns of token cost but does not explain the trade-off or provide decision criteria.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2025-06-18+ | v2 |
No pagination or result limiting documented. Tool accepts 'num' parameter (1-100) but no guidance on default behavior, whether results are paginated, or how to handle large requests.