DSH(DeepSeek Harness)Web GUI 插件集:通知、provider 用量、局域网代理、MCP 管理、闲置归档、文件预览——一键装全家桶或按需单装
This server has 6 tools with significant definition gaps. Tool names use a 'ws_' prefix convention but lack clear action verbs (e.g., 'ws_request_verdict' vs 'request_verdict'). Descriptions are present but brief (18-80 chars), below the 50-200 char baseline for LLM optimization. Input schemas are visible for all tools, but most parameters lack descriptions, only 'presetId', 'sessionId', 'query', 'serverId', 'toolName', and 'arguments' are documented. No output schemas are documented. Error handling is absent, no recovery guidance, categorization, or actionable error messages. The server acts as a middleware/orchestrator for MCP tools and Jev decisions, but tool composition and chaining are not well-documented.
List available Jev verdict guides and presets
Call an MCP tool through the middleware layer
Get detailed information about a specific MCP tool
List all available MCP tools from connected servers
Search MCP tools across connected servers
Request a calibrated Jev decision (Choice/Score/Noul) with local secret precheck and dual-track keys
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields without knowing response structure.
Tool descriptions are too brief (18-80 chars, below 50-200 baseline). Descriptions lack WHEN to use, prerequisites, and state-modification warnings. E.g., 'ws_request_verdict' does not explain what 'calibrated Jev decision' means or when to call it vs alternatives.
No error handling or recovery guidance. Tools do not document what errors are possible, whether they are retryable, or what the LLM should do next. E.g., ws_mcp_call does not explain what happens if the tool fails or the server is unreachable.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 42 | 2026-07-28+ | v2 |
Tool names use 'ws_' prefix (likely WebSocket-related) but lack clear action verbs. 'ws_request_verdict' is acceptable, but 'ws_list_verdict_guides' and 'ws_mcp_list' are generic. Names should start with verb_noun (e.g., 'list_verdict_guides', 'list_mcp_tools').
Parameter descriptions are missing or minimal. 'arguments' in ws_mcp_call is described as 'Tool arguments' but does not explain format, constraints, or how to construct it. LLMs cannot infer valid argument structures.