Loading ToolBench
MCP server for Xiaohongshu (Little Red Book) content publishing, discovery, interaction, and analytics. Provides tools for login, content management, feed interaction, user profiles, and analytics.
Scoring was not performed
NO INPUT SCHEMAS VISIBLE: No tool registration code shown in provided source. Cannot verify JSON Schema input definitions for any of 26 tools. Tool definitions appear to be in InitMCPServer() which is not included. Hard cap on schema scores to 0 per HARD SCORING RULE.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 23 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 25 | - | v1 |
MINIMAL DESCRIPTIONS: Tool descriptions are between 5-20 words and provide only bare minimum context. Example: 'Check the current login status and retrieve user information' (9 words). No explanation of WHEN to use, WHAT parameters are expected, or WHAT the output contains. Violates pattern:tool-description which requires 10-1024 char descriptions with WHEN/WHY/WHAT details.
NO PARAMETER DOCUMENTATION: Descriptions do not document required/optional parameters, their types, allowed values, ranges, or formats. Example: 'publish' tool description does not explain what parameters it expects (text? images? title?). This forces LLMs to guess and violates pattern:constrained-input and pattern:tool-description.
NO OUTPUT SCHEMA DOCUMENTATION: Descriptions do not specify what fields are returned or their types. Example: 'get_feed_detail' does not document what properties of a feed are returned. No pagination hints, no field names, no types. Violates pattern:response-shaper and pattern:tool which require documented return types.
DESTRUCTIVE OPERATIONS LACK SAFETY GUARDS: Tools 'delete_feed', 'delete_comment', and 'delete_cookies' perform irreversible actions but have no visible dry-run, confirmation, or recovery guidance in descriptions. Violates pattern:confirmation-request and pattern:command-tool which require state-modifying tools to have safety patterns.
STATE-MODIFYING OPERATIONS NOT CLEARLY MARKED: Write tools (publish, save_draft, post_comment, like_feed, etc.) lack descriptions stating that they modify state and are not safely retryable. LLMs need to know which tools have side effects. Violates pattern:command-tool.
NO ERROR HANDLING GUIDANCE: Descriptions do not explain error conditions, recovery steps, or when to retry. Example: 'publish' tool has no notes about what happens if the user is not logged in, the image fails to upload, or the network times out. Violates pattern:recovery-guide which requires tools to guide LLM error handling.
PAGINATION NOT DOCUMENTED: Tools 'list_feeds' and 'search_feeds' return potentially large result sets but descriptions do not mention pagination, limit, or offset parameters. Violates pattern:paginated-result which requires discovery tools to support offset/limit and document result counts.
TOOL COMPOSITION BROKEN: Multiple related tools exist but descriptions do not clarify workflow dependencies. Example: 'check_login_status', 'get_login_qrcode', and 'sync_cookies' are authentication tools but descriptions do not explain the intended workflow or when to call each. LLM must guess the sequence. Violates pattern:tool-chain.
TOOL DEFINITIONS INFERRED NOT VISIBLE: Only tool names and brief descriptions are shown in the provided source. The actual tool registration code (likely in InitMCPServer()) is not included. Per HARD SCORING RULE: 'If you are inferring tool existence (not seeing explicit registration): cap that tool's overall at 50.' All tools are capped at 50 individually, but schema=0 applies globally.