Skill flagged — suspicious patterns detected

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

Context Manager

v1.1.0

[何时使用]当用户需要管理个人上下文(Context)时;当用户说"管理我的上下文""整理重要信息""记录关键认知"时;当检测到"决策背景""记忆外脑""认知地图""内心认知"等关键词时;基于 AI 时代个人上下文管理需求,打造思维树构建师

0· 107·0 current·0 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 lj22503/context-manager-v1.

Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Context Manager" (lj22503/context-manager-v1) from ClawHub.
Skill page: https://clawhub.ai/lj22503/context-manager-v1
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 context-manager-v1

ClawHub CLI

Package manager switcher

npx clawhub@latest install context-manager-v1
Security Scan
Capability signals
CryptoRequires sensitive credentials
These labels describe what authority the skill may exercise. They are separate from suspicious or malicious moderation verdicts.
VirusTotalVirusTotal
Benign
View report →
OpenClawOpenClaw
Suspicious
medium confidence
Purpose & Capability
Overall the code aligns with the stated purpose (collecting user notes, tagging, building thought-trees, maps, local storage). However the SKILL/README repeatedly claim integration with external platforms (Feishu, WeChat, Xiaohongshu) and demonstrate commands that imply online API calls; the provided collector implementation either expects exported text (wechat) or stores a placeholder and a source_url for feishu (no actual HTTP fetch implemented). That mismatch between advertised capability and implemented behavior is a coherence issue (not necessarily malicious) and may mislead users about what the skill will access.
Instruction Scope
SKILL.md instructs workflows that collect from multiple sources and shows examples of using doc tokens, but the instructions do not ask the agent to read unrelated system files or to transmit data to external endpoints. The runtime code writes many JSON files to user-controlled base_path (~~/kb or ~/context). The instructions mention API permissions in troubleshooting (e.g., '微信/飞书 API 权限?') which could prompt an agent or user to supply tokens even though the skill declares no required env vars.
Install Mechanism
No install spec is provided (instruction-only in registry), and there are no downloads or brew/npm installs. The code bundle exists in the repository but nothing in the provided files indicates remote code fetches or extract-from-URL installs. This is low-risk from an installer perspective.
Credentials
The skill declares no required environment variables or credentials. Some modules reference a config key (e.g., AIAnalyzer.api_key / ai_model) and README/examples show passing Feishu doc tokens; those are optional config values rather than required env vars. This is proportionate, but be aware that API keys or tokens would live in config.yaml/base_path if you supply them — the skill does not declare or enforce secure storage.
!
Persistence & Privilege
The code creates and writes many files under a default base_path (config defaults like '~/kb' or '~/context'). It will create directories (inbox, contexts, bridges, maps, logs) and persist user content and metadata locally. While expected for a knowledge manager, this persistent file I/O means the skill will store potentially sensitive personal thoughts/decisions on disk by default; you should verify and, if necessary, change the base_path to a controlled location and review file permissions before use.
What to consider before installing
Things to check before installing or running this skill: - Capability vs implementation: The docs promise fetching from Feishu/WeChat/Xiaohongshu; the code accepts tokens/exports but does not implement network fetching for Feishu (it stores a placeholder/URL). If you expect automatic remote fetching, verify/implement the API client and audit any added network calls. - Local persistence: The skill will create directories and write JSON files in a default directory (~/kb or ~/context). These files may contain private inner-cognition entries, decision history, and any tokens/URLs you pass. Change base_path in config.yaml to a safe location and review file contents and permissions. - Secrets handling: The code references an api_key field in config for AI calls but the skill declares no required env vars. If you add API keys/tokens to config.yaml, understand they are stored in plain text. Prefer a secure secret store or environment variables you control and audit any code paths that use them. - Network & omitted files: The provided files (truncated in manifest) show no obvious exfiltration or remote endpoints in the excerpts. However not all files were fully shown; review the remaining files for HTTP requests, subprocess execution, or hidden endpoints before trusting the skill. - Run in sandbox first: Execute the skill in a sandbox or test environment, with a non-sensitive base_path and without providing real API credentials, to confirm behavior. - If you plan to use remote integrations (Feishu, WeChat), require explicit, documented code that performs network calls; otherwise treat the skill as a local organizer that needs manual exports. If you want, I can scan the omitted files for network calls, subprocess usage, or secret-access patterns and flag anything suspicious.

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

ai-assistvk9729b2yc0fe9kzxneryee0r1584yzx4cognitivevk9729b2yc0fe9kzxneryee0r1584yzx4contextvk9729b2yc0fe9kzxneryee0r1584yzx4latestvk9729b2yc0fe9kzxneryee0r1584yzx4memoryvk9729b2yc0fe9kzxneryee0r1584yzx4
107downloads
0stars
1versions
Updated 1w ago
v1.1.0
MIT-0

上下文管理师 🧠

AI 驱动的系统性问题分析与解决方案生成工具

让知识从"收藏"变成"认知"


📋 功能描述

基于 Karpathy LLM Wiki + 上下文管理理论,帮助用户在 AI 时代管理个人上下文。

核心洞察

知识管理的本质不是管理知识,而是管理个人上下文(Context)。

上下文 = 个人知识 + 经历 + 情感 + 判断 = 认知框架

适用场景

  • 管理微信/飞书/小红书等分散信息
  • 记录内心认知(触动事 + 判断)
  • 构建思维树(自动关联 + 认知地图)
  • 决策时快速回溯相关上下文
  • 每周回顾认知更新
  • 设置 AI 辅助边界

边界条件

  • 不替代深度思考(明确 AI 辅助边界)
  • 需配合人工标注意义标签
  • 内心认知需要主动记录

⚠️ 常见错误

错误 1:过度依赖 AI

问题:
• 所有思考都交给 AI,弱化自主思考能力

解决:
✓ 明确 AI 辅助边界:哪些事必须自己思考
✓ AI 生成摘要,但最终决策手动确认
✓ 保留"不记录的权利",给直觉留空间

错误 2:信息过载

问题:
• 每天接触大量碎片内容,难以筛选真正重要的

解决:
✓ 设置关键词触发("重要""决定""灵感")
✓ 每天只记录 3 件触动事
✓ 每季度删除冗余内容

错误 3:工具割裂

问题:
• 微信/飞书/小红书数据无法互通

解决:
✓ 统一导入到上下文管理库
✓ 用 AI 做"记忆外脑",自动整理
✓ 建立"上下文桥接机制"

错误 4:认知私有化

问题:
• 内心感受和判断难以用语言准确记录

解决:
✓ 用语音或文字快速记录(哪怕不完整)
✓ 建立"个人意义标签体系"
✓ 每周生成"认知地图"

🧪 使用示例

示例 1:记录内心认知

输入

@ant context-manager:
类型:内心认知日志
内容:今天和朋友争论时,我发现自己忘记了曾经也有过类似经历。
触动点:记忆断层
我的判断:需要建立"经历回溯"机制

输出

---
type: 内心认知日志
date: 2026-04-14
tags: [#记忆断层 #经历回溯 #关系锚点]
---

# 内心认知日志

**触动事**:今天和朋友争论时,我发现自己忘记了曾经也有过类似经历。

**触动点**:记忆断层

**我的判断**:需要建立"经历回溯"机制

**AI 分析**:
- 情绪:遗憾 + 觉察
- 模式:第 3 次出现类似情况
- 建议:建立"争论前回溯"习惯

**关联上下文**:
- [[2026-03-15 类似争论]]
- [[记忆管理策略]]

示例 2:构建思维树

输入

@ant context-manager build-tree:
主题:知识管理
时间范围:最近 3 个月

输出

# 思维树:知识管理(2026-01 至 2026-04)

## 核心认知
- 知识管理的本质是管理个人上下文
- 不是"存",而是"用"
- 让知识自动生长

## 外部信息
- 飞书文档《如何建立知识管理体系》
- 微信读书《价值》划线
- 小红书笔记"知识管理痛点"

## 内心认知
- 2026-02-10:收藏 1000 篇,读完<10% → 痛点
- 2026-03-15:知识库变"数字坟墓" → 反思
- 2026-04-14:知识管理=上下文管理 → 洞察

## 上下文桥接
微信收藏 → 引发痛点 → 飞书文档 → 形成方案 → 小红书验证

## 认知更新
- 旧认知:知识管理=整理
- 新认知:知识管理=上下文管理

示例 3:决策回溯

输入

@ant context-manager:
类型:决策回溯
决策:要不要加仓这只基金?

输出

# 决策回溯:要不要加仓这只基金?

## 相关上下文(快速回溯)

### 1. 投资原则(2026-01-10)
- 资产配置纪律:股债 60/40
- 单只基金上限:10%
- 不加仓理由:保持纪律

### 2. 类似决策(2025-12-15)
- 当时决策:不加仓
- 结果:正确(避免了下跌)
- 认知:纪律>短期收益

### 3. 内心认知(2026-04-10)
- 触动点:看到别人赚钱,想追高
- 判断:FOMO 情绪,需要警惕
- 标签:#FOMO #纪律

## AI 建议
- 情绪分析:贪婪 70%,恐惧 30%
- 模式识别:第 5 次出现 FOMO
- 建议:不加仓,保持纪律

## 最终决策
[ ] 加仓  [x] 不加仓  [ ] 观望

**决策理由**:保持资产配置纪律,避免 FOMO 情绪

🔧 故障排查

问题检查项
不触发description 是否包含触发词?
上下文收集失败微信/飞书 API 权限?
思维树为空是否有足够的内心认知记录?
决策回溯无结果是否有历史决策记录?
认知地图混乱是否定期删除冗余内容?

🎯 核心功能

1. 上下文收集

支持来源

  • 微信聊天记录
  • 飞书文档
  • 小红书笔记
  • 内心认知日志
  • 网页文章
  • 手动笔记

使用示例

@ant context-manager collect:
来源:微信
内容:[粘贴聊天记录]

@ant context-manager collect:
来源:内心认知
内容:今天有个洞察...

2. 思维树构建

功能:自动关联外部信息 + 内心认知,形成认知地图。

使用示例

@ant context-manager build-tree:
主题:知识管理
时间范围:最近 3 个月

输出

  • 核心认知
  • 外部信息
  • 内心认知
  • 上下文桥接
  • 认知更新

3. 意义标签体系

标签类型

  • #成长痛点# - 成长过程中的痛点
  • #关系锚点# - 重要关系的锚点事件
  • #灵感触发# - 灵感触发的瞬间
  • #认知冲突# - 认知冲突的时刻
  • #决策背景# - 重要决策的背景

使用示例

@ant context-manager tag:
内容:今天和朋友争论...
标签:#关系锚点# #认知冲突#

4. 认知日志

功能:每天记录 3 件触动事 + 你的判断。

使用示例

@ant context-manager daily-log:
1. 触动事:... 判断:...
2. 触动事:... 判断:...
3. 触动事:... 判断:...

5. 上下文桥接

功能:把内心想法与外部信息用箭头连接。

使用示例

@ant context-manager bridge:
内心想法:知识管理=上下文管理
外部信息:飞书文档《知识管理体系》

输出

内心想法 → 引发洞察 → 飞书文档 → 形成方案

6. AI 辅助边界

功能:明确哪些事必须自己思考,哪些可交给 AI。

使用示例

@ant context-manager ai-boundary:
事项:重大人生决定
边界:必须自己思考

事项:会议纪要
边界:可交给 AI

7. 决策回溯

功能:在关键决策时,快速回溯至少 2 个相关上下文点。

使用示例

@ant context-manager recall:
决策:要不要加仓这只基金?

输出

  • 相关投资原则
  • 类似决策历史
  • 内心认知记录
  • AI 建议

8. 认知更新

功能:每季度删除冗余内容,每年重审核心认知。

使用示例

@ant context-manager review:
周期:季度
操作:删除冗余

@ant context-manager review:
周期:年度
操作:重审核心认知

📊 成功指标

指标目标检查方式
每周回顾能复述出 3 个重要认知更新周日报
决策回溯能快速回溯至少 2 个相关上下文点决策记录
情绪焦虑因信息混乱而下降 30%自我评估
认知完整性内心认知与外部数据形成互补认知地图
AI 辅助边界明确哪些自己思考,哪些交给 AI边界清单

🔗 相关技能

  • knowledge-workflow - 知识管理工作流(前置技能)
  • note-tagger - 笔记打标
  • experience-memory-tracker - 体验记忆追踪

推荐组合

context-manager → knowledge-workflow
(上下文管理)    (知识生产)

维护者:燃冰 & ant
版本:v1.0
创建日期:2026-04-14
发布状态:待发布

Comments

Loading comments...