This server has significant structural issues. Tool naming is inconsistent (parse_* prefix dominates, but lacks action-verb clarity for some tools like 'url'). Descriptions vary wildly in quality and length, some are under 50 characters (e.g., 'url' tool: 'Fetches a website and returns its content' at 48 chars), others are longer but lack actionable guidance. Input schemas are present for all visible tools, but parameter descriptions are sparse or missing. Output schemas are not documented. Error handling exists but is generic ('Error: ...'). No evidence of pagination, output limits, or field-level documentation. The 'maxkb' tool appears to be a third-party API wrapper with minimal local context. File path handling is opaque ('process_file_path' method referenced but not shown). Overall, this reads like a collection of convenience wrappers rather than a professionally designed tool suite. Tool composition is reasonable (each does one thing), but descriptions don't guide LLM selection, and output structure is undocumented.
请求MaxKB API并返回处理后的结果
解析CSV文件内容,支持各种编码格式
Parses an Excel file and returns its content including all sheets
解析文件内容,支持PDF、Word、Excel、CSV和Markdown格式
解析Markdown文件内容,提取标题结构、列表和文本内容
解析PDF文件内容,支持快速预览和完整解析两种模式
解析Word文档内容,提取文本、表格和图片信息
Fetches a website and returns its content
Tool 'url' has a non-verb name and critically short description (48 chars). The name alone does not convey intent, 'fetch_url' or 'get_website_content' would be clearer. Description lacks WHEN to use it, prerequisites, or what gets returned.
Parse_file description (43 chars) is too generic and does not explain WHICH file formats are supported or HOW to choose between parse_file vs. parse_pdf vs. parse_csv. LLMs will struggle to select the right tool.
Output schemas are not documented for any tool. CSV output includes 'info', 'stats', formatted strings, but the agent has no schema to guide field extraction. Excel returns JSON with nested 'sheets' object, but structure is undocumented. This violates the output schema pattern.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 0 | 2026-07-28+ | v2 |
| 2026-03-09 | D | 50 | - | v1 |
Maxkb tool lacks meaningful description of what MaxKB is, what this API does, when to call it vs. other tools, and what the 'message' parameter expects. Description is only 36 chars and provides no actionable guidance.
File path handling is opaque. Tools reference 'process_file_path()' method but it is not shown in source. Security concern: path traversal attacks possible if this method does not validate paths. LLM cannot reason about what paths are safe to pass.
Parameter descriptions are minimal or missing. CSV tool 'encoding' param has a description, but most other optional params (e.g., 'stream' in maxkb, 're_chat' in maxkb) lack actionable context. The 'mode' enum in parse_pdf is documented, but other params are not.
Error messages are generic strings like 'Error: Missing required argument' and 'Error: File not found'. No guidance on recovery (e.g., 'Try passing a valid file path' or 'Call list_files() to see available files'). Error classification (retryable vs. user-fixable) is absent.
Parse_excel returns entire DataFrame as JSON with all rows ('data': sheet_data). For large Excel files, this could exceed context limits. No pagination, no result limits, no cardinality documentation. Baseline: A+ tools cap results at 20-50 items.
Parse_csv and parse_excel return preview/summary data, but structure is inconsistent. CSV returns formatted strings with stats; Excel returns JSON with structured 'sheets'. No documented output schema. Agent cannot reliably parse responses.
URL tool has no content-length limit, no timeout declared, no guidance on when to use it vs. file parsing tools. Fetching a 1GB binary file would fail silently or hang. No safeguards against redirect loops or malicious servers.