Skill flagged — suspicious patterns detected

ClawHub Security flagged this skill as suspicious. Review the scan results before using.

Smart Todo - AI智能代办管理

v1.0.0

智能代办管理 Skill。用于管理日常任务和代办事项,支持优先级标记(P0-P1)、 状态跟踪、智能重复检测、上下文保存、定时提醒等功能。 当用户需要添加、查看、更新代办,或检测到工作中断时自动使用。

0· 93·1 current·1 all-time

Install

OpenClaw Prompt Flow

Install with OpenClaw

Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for 13770626440/smart-todo.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Smart Todo - AI智能代办管理" (13770626440/smart-todo) from ClawHub.
Skill page: https://clawhub.ai/13770626440/smart-todo
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.

Command Line

CLI Commands

Use the direct CLI path if you want to install manually and keep every step visible.

OpenClaw CLI

Bare skill slug

openclaw skills install smart-todo

ClawHub CLI

Package manager switcher

npx clawhub@latest install smart-todo
Security Scan
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
The declared purpose (todo management, context capture, duplicate detection, reminders) is consistent with the included Python modules. The code implements scanning of the workspace for recently modified files, reading .workbuddy/memory markdown files, capturing environment info, and writing markdown todo files to a storage_path. Minor inconsistency: SKILL.md examples show a Windows path (D:\knowledge\代办) while assets/config.json defaults to a home-directory path (~/smart-todo-data). This is likely a documentation/config mismatch but not by itself malicious.
!
Instruction Scope
SKILL.md explicitly instructs the agent to capture '当前打开的文件、最近的文件修改、对话历史(最近10条)' and the code implements workspace-wide os.walk scans (limited to certain file types and counts) and reading .workbuddy/memory markdown files. Critically, SKILL.md states that if the user refuses creating a todo during an interruption, the skill will '仍静默保存上下文供后续参考' (still silently save the context). That silent-save behavior is scope-creep/privacy-sensitive and not something most users would expect from a todo helper.
Install Mechanism
No install spec is provided (instruction-only + included scripts). Nothing is downloaded or executed from external URLs during install. This is the lower-risk pattern compared to remote installers.
Credentials
The skill requests no environment variables or external credentials. However, it reads local filesystem contents (recent files, .workbuddy memory files), captures current working directory and writes files under storage_path (default ~/smart-todo-data). Access to arbitrary files in the user's workspace is proportionate to an auto 'context-capture' feature but potentially exposes sensitive local data; the lack of explicit consent for silent saves increases disproportion.
!
Persistence & Privilege
The skill will create/modify files under storage_path (active.md, archive.md) — expected for a todo manager. It does not request 'always: true', but the platform default allows autonomous invocation. Combined with the SKILL.md claim that it may automatically capture and silently save context during detected interruptions, this increases the privacy blast radius because the agent could run this skill and store local context even when the user declines.
What to consider before installing
What to consider before installing: - Privacy: The skill scans your workspace for recently modified files and reads .workbuddy/memory/*.md to extract paths and conversation snippets. If your workspace contains secrets or private files, those could be captured into todo context. - Silent saving: SKILL.md explicitly says it will silently save context even if the user refuses creating a todo. If you want explicit consent control, edit SKILL.md or the code before enabling automatic triggers. - Storage location: assets/config.json defaults to ~/smart-todo-data. Change storage_path to a dedicated directory you control (preferably not a repo root or a directory containing secrets) before first run. - Audit & sandbox: Review scripts/context_capture.py and scripts/todo_manager.py line-by-line (they are included) or run the skill in a sandboxed environment first. Look for any places you’d like to disable (e.g., reading .workbuddy memory or os.walk behavior). - Configuration changes: Consider lowering or disabling automatic 'work interruption' triggers, or modify the flow so that nothing is saved unless the user explicitly confirms. Also limit how many files are read or remove heuristics that parse other markdown files. - Backup: Because the skill will create and overwrite active.md/archive.md, back up any existing files that might be overwritten. If you want, I can point out exact lines in the two Python files that implement the workspace scan, .workbuddy memory read, and the 'silent save' behavior so you can edit them safely.

Like a lobster shell, security has layers — review code before you run it.

duplicate-detectionvk9785cmmzwy63c4z7ds4tc840d83yq5ylatestvk9785cmmzwy63c4z7ds4tc840d83yq5yproductivityvk9785cmmzwy63c4z7ds4tc840d83yq5ytask-managementvk9785cmmzwy63c4z7ds4tc840d83yq5ytodovk9785cmmzwy63c4z7ds4tc840d83yq5y
93downloads
0stars
1versions
Updated 3w ago
v1.0.0
MIT-0

Smart Todo Manager

功能概述

智能代办管理系统,提供以下核心能力:

功能说明
优先级管理P0(最高)、P1(普通)
状态跟踪未开始、进行中、已完成、暂停、终止
智能重复检测名称+描述双维度,80%相似度阈值
上下文保存自动捕获工作文件、对话历史、任务状态
定时提醒20分钟间隔状态汇报,P0超时提醒
自动归档已完成/终止代办自动转移到归档文件

触发条件

显式触发

用户明确表达添加代办意图:

  • "加入代办:修复登录bug"
  • "记一下,明天要开会"
  • "添加到待办"
  • "创建代办任务"

查询触发

用户询问代办状态:

  • "查看我的代办"
  • "待办列表"
  • "有什么待办"
  • "P0 代办有哪些"

工作中断触发

检测到用户切换任务意图:

  • "先放下这个"
  • "换个事"
  • "等一下"
  • "先做别的"
  • "转移注意力"

中断处理流程

  1. 捕获当前任务上下文(打开文件、对话历史、任务状态)
  2. 询问用户:"检测到任务切换,是否为当前任务创建代办?"
  3. 如确认,创建包含完整上下文的代办
  4. 如拒绝,仍静默保存上下文供后续参考

工作流程

流程1:添加代办

1. 解析用户输入
   └─ 提取任务核心内容
   
2. 生成简略名称(≤10字)
   └─ 示例:"修复登录API超时问题" → "修复登录超时"
   
3. 确认理解
   └─ 与用户对话确认任务理解正确
   └─ 记录确认后的理解内容
   
4. 捕获上下文
   └─ 当前打开的文件
   └─ 最近的文件修改
   └─ 对话历史(最近10条)
   └─ 当前任务目标
   
5. 重复检测
   └─ 名称相似度 >=70% 或 描述相似度 >=80%?
   └─ 任一满足 → 触发重复警告
   
6. 用户确认
   ├─ 无重复 → 直接创建
   └─ 有重复 → 询问:合并/新建/取消
   
7. 保存代办
   └─ 写入 active.md
   └─ 更新统计信息

流程2:重复检测

检测逻辑(OR 组合,任一维度触发即报警):

if similarity(name_new, name_existing) >= 0.7 or \
   similarity(desc_new, desc_existing) >= 0.8:
    trigger_duplicate_warning()

名称阈值 70%(中文相似词匹配较松)、描述阈值 80% 均可在 config.json 中调整

用户提示

⚠️ 检测到相似代办:

现有代办 [T003] P0 修复登录超时
- 描述: 用户登录API返回超时,需优化数据库查询
- 相似度: 名称 85%, 描述 90%

你的新代办是否与此相同?
[是,合并] [否,新建] [查看详情] [取消]

流程3:状态更新

状态流转图

未开始 ──→ 进行中 ──→ 已完成 ✓ (归档)
   ↑         ↓
   └──── 暂停 ─┘
              ↓
            终止 ✓ (归档)

归档规则

  • 状态变为"已完成" → 移动到 archive.md,记录完成时间
  • 状态变为"终止" → 移动到 archive.md,记录终止原因

流程4:定时提醒

触发时机:每次会话开始时检查

提醒内容

📋 代办状态更新 (距上次 25 分钟)

活动代办: 5 个
├─ P0: 2 个 (⚠️ 1个已超时)
└─ P1: 3 个

⚠️ P0 代办提醒:
[T001] 修复支付接口 - 进行中 (已 1.5 小时)
建议: 是否需要帮助或调整优先级?

快捷操作:
- "标记 T001 完成"
- "暂停 T001"
- "查看详情"

文件结构

D:\knowledge\代办\
├── active.md           # 活动代办
├── archive.md          # 归档代办
└── config.json         # 用户配置(可选覆盖)

active.md 格式

# 活动代办列表

## 统计
- 总计: 3
- P0: 1
- P1: 2
- 进行中: 1
- 未开始: 1
- 暂停: 1

## 代办列表

### [T001] P0 修复登录超时
**状态**: 进行中
**创建时间**: 2026-03-31 10:00:00
**更新时间**: 2026-03-31 14:30:00
**预估时间**: 60分钟

**原始描述**:
用户反馈登录经常超时,需要排查是数据库问题还是网络问题

**确认理解**:
优化登录接口性能,目标响应时间 < 500ms

**上下文**:
捕获时间: 2026-03-31 10:05:00

【当前目标】
优化登录接口性能

【打开的文件】
  - src/api/auth.py
  - src/db/connection.py

【最近文件】
  - config/database.yml
  - tests/test_auth.py

【对话摘要】
  > 用户: 登录超时问题需要处理
  > 助手: 已记录,建议先检查数据库连接池

【备注】
工作中断 | 原因: 需要紧急处理支付bug

**标签**: #bug #性能优化

---

### [T002] P1 编写API文档
...

---
最后更新: 2026-03-31 15:00:00

配置选项

assets/config.json

{
  "storage_path": "D:\\knowledge\\代办",
  "notification_interval_minutes": 20,
  "similarity_threshold": 0.8,
  "max_name_length": 10,
  "context_lines_from_history": 10,
  "p0_alert_threshold_hours": 1
}
配置项说明默认值
storage_path代办文件存储路径D:\knowledge\代办
notification_interval_minutes提醒间隔(分钟)20
similarity_threshold重复检测阈值0.8
max_name_length简略名称最大长度10
context_lines_from_history保留对话历史条数10
p0_alert_threshold_hoursP0超时提醒阈值(小时)1

使用示例

添加代办

用户: "加入代办,修复登录超时问题,用户反馈登录经常卡住"

助手:

已理解,创建代办:

简略名称: 修复登录超时 (8字)
优先级: P1 (默认)
预估时间: 20分钟 (默认)

确认理解:
用户登录接口响应慢,需要排查性能问题并优化

是否正确?
[正确,保存] [修改名称] [修改优先级] [取消]

检测到重复

用户: "加入代办,修复登录问题"

助手:

⚠️ 检测到相似代办:

现有代办 [T001] P0 修复登录超时
- 描述: 用户反馈登录经常超时...
- 相似度: 名称 90%, 描述 85%

操作选项:
1. 合并 - 更新现有代办
2. 新建 - 这是不同的问题
3. 查看 - 显示现有代办详情
4. 取消

工作中断检测

用户: "先放下这个,我去处理支付bug"

助手:

检测到任务切换。

当前任务: 修复登录超时 (进行中)
已耗时: 30分钟
打开文件: src/api/auth.py, src/db/connection.py

是否为当前任务创建代办以保存进度?
[是,保存上下文] [否,继续] [稍后提醒]

查看代办

用户: "查看代办"

助手:

📋 代办列表 (3个活动)

P0 (1):
├─ [T001] 修复登录超时 - 进行中 (30分钟)

P1 (2):
├─ [T002] 编写API文档 - 未开始
└─ [T003] 更新依赖包 - 暂停

操作: "查看 T001" / "完成 T002" / "全部详情"

边界情况处理

场景处理方式
名称超过10字自动截断并提示用户
重复检测冲突显示相似代办,等待用户选择
P0超时未更新主动提醒,询问是否需要帮助
上下文捕获失败记录失败原因,继续创建代办
存储路径不存在自动创建目录
文件解析错误备份原文件,重新初始化

实现依赖

  • scripts/todo_manager.py - 核心CRUD和重复检测
  • scripts/context_capture.py - 上下文捕获
  • assets/config.json - 配置
  • references/todo_template.md - 模板参考

版本

v1.0.0 - 初始版本

  • 基础CRUD功能
  • 重复检测
  • 上下文捕获
  • 自动归档

Comments

Loading comments...