MCP server that reads DeepSeek recharge balance and today's consumption. Data is collected by a local helper process with configurable refresh intervals.
Single tool 'deepseek_usage' has critical definition gaps. Input schema is empty object with no parameters, violates pattern:tool-description and pattern:constrained-input. Description is in Chinese (non-English) and lacks actionable context for LLM selection. No output schema documented. Tool appears to be a read-only status check but lacks error handling guidance, retry semantics, or field documentation. No parameter descriptions (none exist). Naming is acceptable (verb_noun form: 'deepseek_usage') but description quality is severely deficient.
读取 DeepSeek 充值余额与今日消费。数据由本地助手进程采集,刷新间隔见 config.json。
Input schema is empty object with no parameters and no descriptions. Violates pattern:tool-description and pattern:constrained-input. LLM cannot infer what the tool does or when to call it.
Tool description is in Chinese ('读取 DeepSeek 充值余额与今日消费。数据由本地助手进程采集,刷新间隔见 config.json。') and lacks English translation. Non-English descriptions prevent LLM selection in English-language agents. Description is also vague about output structure and error conditions.
No output schema documented. LLM cannot plan downstream operations or extract required fields. Based on UI code, tool returns {toppedUpBalance, todaySpend, symbol, updatedAt, error, ok, todaySpendUnavailable} but this is inferred, not declared.
No error handling guidance. Tool can fail (data.error && !data.ok) but description does not explain error conditions, recovery steps, or retry semantics. LLM has no guidance on what to do if the helper process is unavailable.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 33 | 2026-07-28+ | v2 |
Tool depends on external 'helper process' and config.json refresh interval, but these dependencies are not documented in the tool description. LLM cannot understand prerequisites or failure modes.