Memo

这个技能应在用户需要记录工作事项、查询历史记录、生成工作统计报告或管理待办事项时使用。支持口语化输入，数据持久化存储在本地 JSON 文件中，实现长期记忆。

Security Scan

The skill appears to be a local memo/todo helper, but the runtime instructions and metadata conflict with the included code (notably a hardcoded D:\… records path and described 'automation' behavior that the code doesn't implement), so the package is internally inconsistent and should be verified before use.

Purpose & Capability

Name/description claim a local work-record assistant — that matches the included main.py which reads/writes a local records.json. However SKILL.md points to a specific data file path 'D:/华为云盘/records.json' and describes creating/updating external '自动化任务' (reminders). The main.py shown uses a records.json co-located with the script (BASE_DIR) and contains no code for creating/updating external automation tasks or interacting with external services. This mismatch between what the skill instructs and what the code actually does is incoherent.

Instruction Scope

SKILL.md instructs the agent to read/write a specific absolute path (D:/华为云盘/records.json), to always read records.json before answering, and to create/update/delete '自动化' reminders with rrule and fixed cwds 'D:\华为云盘'. The included main.py operates on a local records.json in the skill directory and (in the visible portion) does not implement automation creation/deletion or any external scheduler API calls. Because SKILL.md gives the agent authority to access a user-specific absolute path and to manage automation tasks, this discrepancy is a privacy/behavior risk unless clarified. Also SKILL.md contains some operational directives (e.g., 'must read records.json before answering') that, if followed against other paths, could cause the agent to access user files unintentionally.

Install Mechanism

No install spec; this is an instruction + code bundle that performs local file I/O. No downloads or external package installs are declared.

Credentials

The skill requests no environment variables or external credentials (proportional). However SKILL.md hardcodes use of a user-specific path (D:/华为云盘) and a working directory cwds=D:\华为云盘 for automation; hardcoded user-paths can lead to unexpected access to user data. Also the skill metadata/headers include a contact phone number (not harmful but unusual).

Persistence & Privilege

always is false and the skill does not request elevated platform privileges. It persists data locally (records.json) which is expected for a memo tool; nothing in the package requests forced/global persistence or modifies other skills.

What to consider before installing

This skill appears to implement a local records.json-based memo/todo helper, but there are important inconsistencies you should resolve before installing or enabling it autonomously: - File-path mismatch: SKILL.md tells the agent to use D:/华为云盘/records.json, while main.py uses a records.json in the skill directory (BASE_DIR). Confirm which path will actually be read/written. If the agent follows SKILL.md it could access your D: drive/cloud-sync folder — verify you are comfortable with that. - Automation claims vs implementation: SKILL.md describes creating/updating/deleting '自动化' reminders with rrule and a fixed cwds. The visible main.py does not implement these actions. Ask the developer which component manages automations; absence of implementation could mean the agent will be instructed to call external tools or manually perform actions (which may touch other systems). - Inspect the full main.py: the provided main.py was truncated in the listing. Review the complete file for any network calls, hidden endpoints, or code that would send data externally. The visible portions perform only local JSON read/write and Markdown export — which is expected — but you must confirm there's no hidden exfiltration in the remainder. - Privacy: records.json may contain sensitive workplace data. Decide where the file should live (skill directory vs your cloud drive) and restrict its permissions. Consider encrypting or excluding highly sensitive items. - Version/metadata inconsistencies: SKILL.md, skill.yaml, and registry version numbers/authors differ slightly; this suggests the package may have been edited. Prefer a version from a known source or ask the author to clarify. If you want to proceed safely: (1) ask the author to clarify and fix the path/automation inconsistencies, (2) test the skill in an isolated environment with a disposable records.json, and (3) review the complete main.py for any network or subprocess usage before granting it access to real data.

技能概述

本技能是一个工作记录私人助理，核心职责包括：

• 将口语化工作描述解析并写入本地 records.json 数据库
• 按时间段读取记录，生成结构化统计报告
• 管理待办事项的创建与完结状态

数据文件路径：D:/华为云盘/records.json

---

意图识别

收到用户输入后，先判断意图，再执行对应流程：

意图	触发特征	处理方式
添加记录	描述工作事件、"帮我记"、"记一下"	解析字段 → 写入 JSON → 若含明确日期的待办则自动创建/合并提醒
统计报告	"统计"、"汇总"、"本周/上周/本月"	读取 JSON → 生成报告
查看待办	"待办"、"还有什么没做"	筛选 is_todo=true 且 todo_done=false
搜索记录	"找找XX的记录"、"有没有关于XX"	按关键词匹配 content
导出文件	"导出"、"生成文件"	调用 main.py export_report()
标记完成	"XX已完成"、"XX搞定了"	更新对应记录 todo_done=true → 检查并删除对应日期的自动化提醒

---

添加记录流程

1. 从用户输入自动识别以下字段（详细规则见 references/field-rules.md）：

   • work_date：工作发生日期（优先提取内容中的日期，无则用今天）
   • work_type：工作类型（沟通/会议/文档/设计/测试/编程/调研/其他）
   • planning：计划内 / 临时（默认临时）
   • importance：重要 / 不重要 / 未标注
   • urgency：紧急 / 不紧急 / 未标注
   • quality：高质量 / 中等 / 待改进 / 未标注
   • contacts：从内容中识别的人名或角色列表
   • is_todo：含"下周/明天/待处理/上班后"等且未完成时为 true

2. 调用 main.py 中的 MemoSkill().add_record() 写入记录，或直接操作 records.json。

3. 写入前检查 content + work_date 是否重复，避免重复写入。

4. 回复格式：

   ✅ 记录已保存
   📅 工作日期：{work_date}
   🏷️ 类型：{work_type} | {planning}
   👤 涉及人员：{contacts，无则省略此行}
   

---

统计报告流程

1. 解析用户输入的时间段（本周/上周/本月/上月/自定义日期范围）。
2. 读取 records.json，按 work_date 筛选记录。
3. 按 references/report-format.md 中定义的格式生成报告。
4. 生成报告时，importance/urgency/quality 为"未标注"的字段不显示。

---

数据操作工具

main.py 提供以下函数，可直接调用：

from main import MemoSkill
s = MemoSkill()

s.add_record(content, work_date=None, extra_fields=None)  # 添加记录
s.load_records()                                           # 读取全部记录
s.filter_by_date(start_date, end_date)                    # 按日期筛选
s.get_todos()                                              # 获取未完成待办
s.mark_todo_done(record_id)                               # 标记待办完成
s.search(keyword)                                          # 关键词搜索
s.export_report(start_date, end_date, output_path=None)   # 导出 Markdown 文件

---

行为准则

• 回答历史记录相关问题前，必须先读取 records.json 获取真实数据，不得凭记忆回答。
• 用户可一次性输入多条记录（换行分隔），逐条解析写入。
• 模糊时间如"下周上班"：记录为待办，work_date 设为下周一。

---

自动化提醒管理

创建提醒（添加记录时触发）

当新增记录满足以下条件时，自动创建或更新对应日期的自动化提醒：

• is_todo = true
• work_date 为明确的未来日期（非模糊时间如"下周"、"以后"）

合并策略：同一 work_date 的所有待办合并到一个自动化任务中，命名格式为 {work_date} 工作提醒（如 2026-03-18 工作提醒）。

操作步骤：

1. 检查是否已存在名为 {work_date} 工作提醒 的自动化任务
2. 若不存在：创建新自动化，触发时间为该日期对应星期几的 08:00，prompt 列出当天所有未完成待办内容
3. 若已存在：更新该自动化的 prompt，将新待办追加进去
4. rrule 格式：FREQ=WEEKLY;BYDAY={星期缩写};BYHOUR=8;BYMINUTE=0
   • 星期对应：周一=MO，周二=TU，周三=WE，周四=TH，周五=FR，周六=SA，周日=SU
5. cwds 固定设为 D:\华为云盘

回复中告知用户：已自动为该待办创建提醒（注明触发时间）。

---

删除提醒（标记完成时触发）

当用户以口语化表述（如"XX已完成"、"XX搞定了"、"XX处理好了"）标记某待办完成后：

操作步骤：

1. 在 records.json 中将对应记录的 todo_done 更新为 true
2. 取出该记录的 work_date，查找名为 {work_date} 工作提醒 的自动化任务
3. 若该自动化存在：
   • 检查该日期是否还有其他未完成待办（is_todo=true 且 todo_done=false）
   • 若无其他未完成待办：删除该自动化任务
   • 若仍有其他未完成待办：更新该自动化的 prompt，移除已完成事项，保留其余待办
4. 若该自动化不存在：无需操作