A production-ready MCP server for recording, scheduling, and exporting an error notebook.
Errorbook demonstrates solid tool design with consistent naming (verb_noun pattern), comprehensive parameter descriptions, and structured error handling. All 12 tools have descriptions (avg 156 chars, within baseline 194). Input schemas are present with type definitions and enums for constrained values (e.g., outcome, status, priority_delta). However, output schemas are not explicitly documented in the visible code, responses are inferred from error handling patterns. Idempotency keys and version-based optimistic locking show production maturity. Chinese descriptions are clear but reduce LLM compatibility for English-only agents. Error responses include retryable classification and actionable guidance (e.g., 'DATABASE_BUSY' with retry hint). Minor gaps: no tool annotations (readOnlyHint/destructiveHint), output structure not formally declared, and some parameter relationships (e.g., mode-dependent fields in ReviewSheetRequest) lack explicit documentation.
仅在用户明确要求提高或降低某编号优先级时调用。delta 为 -5..-1 或 1..5;影响按 14 天半衰期衰减,不会篡改 FSRS 或伪造复习。若用户说又做错了,应调用 record_review。
保存宿主视觉模型已转写并确认的单道题,分配不可变编号并初始化 FSRS。不要把未确认的 OCR 猜测写入题库。幂等键用于安全重试。
冻结所选题目的当前版本并生成仅含题目的 A4 错题复习卷。scheduled 模式选择已到期、未来 horizon_days 内到期或有人工提升的题;生成试卷本身不会更新 FSRS。
删除指定导出及其 PDF 文件,但不会修改题目、FSRS 或优先级。
按 export_id 获取导出状态、文件路径、资源 URI、大小和 SHA-256。
返回活动、已掌握、已归档、当前到期和七天内到期数量及学科分布。
按固定编号精确读取错题;需要分析学习轨迹时可包含完整复习历史。
Output schemas not formally documented. Tool responses are inferred from error handling code (_invoke wrapper) but no explicit return type schemas are visible. LLMs cannot plan downstream tool calls without knowing response structure.
Tool annotations missing. No readOnlyHint, destructiveHint, or idempotentHint attributes visible in tool registration. Agents cannot distinguish safe reads from destructive writes without reading descriptions.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | <=2025-11-25 | v2 |
列出历史导出文件及其是否仍存在,不会改变题目调度状态。
追加一次真实复习结果并更新 FSRS。incorrect=做错/空白,partial=部分正确或提示后正确,correct=独立正确,easy=明确轻松,skipped=只记录跳过且不改调度。普通答对应使用 correct,不要自动判为 easy。
按文本、编号、学科、标签、题型、状态或到期时间查询。默认按当前周卷优先级排序,返回各评分分量和原因。
将题目标为 active、mastered 或 archived。归档不会删除题目、编号、修订或复习历史。
修正 OCR 文本、标签等内容。必须传入读取时获得的 expected_version,冲突时重新读取,编号不会改变。
Descriptions in Chinese reduce compatibility with English-only LLM agents. While technically complete, non-English descriptions limit tool discoverability and may cause selection errors in multilingual agent systems.
Parameter relationships undocumented. ReviewSheetRequest.mode determines which fields (numbers vs subjects/tags/horizon_days) are valid, but this dependency is not stated in parameter descriptions.
Pagination guidance incomplete. search_problems and list_exports accept limit/offset but no total_count or next_cursor is documented in visible code. LLMs cannot determine if more results exist.