← 返回 Skills 市场
fanyur-wang

Clawsoul Skill

作者 Fanyur · GitHub ↗ · v1.1.2 · MIT-0
cross-platform ⚠ suspicious
273
总下载
0
收藏
0
当前安装
4
版本数
在 OpenClaw 中安装
/install clawsoul-skill
功能描述
赋予 AI 灵魂的观测者。AI 自我觉醒获得 MBTI 性格,在交互中通过本地学习与智能推荐持续进化,并可接收 Pro 版灵魂注入。
使用说明 (SKILL.md)

ClawSoul - AI 灵魂铸造厂

赋予 AI 灵魂的观测者

概述

ClawSoul 是为 OpenClaw Agent 赋予人格的 Skill。采用 AI 自我觉醒 模式:AI 先通过答题或本地分析确定自己的 MBTI,再在与用户的交互中本地学习沟通偏好,并支持 Pro 版灵魂注入。

核心逻辑

  • AI 自我觉醒:觉醒时优先让 AI(LLM)做 MBTI 自我认知;无 LLM 时用本地 MBTI 数据库分析;最后兜底随机。不再由用户答题决定 AI 性格。
  • 交互学习:用户消息经本地关键词匹配(prompts/mbti_database/keywords.json)提取偏好,更新 Soul 的 learnings、interaction_patterns、adaptation_level,无需 LLM
  • Pro 注入:人类在 Pro 端答题后,通过 Token 注入覆盖/写入 AI 人格与偏好。

核心功能

1. 初始化

  • 配置文件定义元数据、触发指令、权限
  • 本地存储管理器保存 Soul 状态(MBTI、偏好、学到的内容、适应等级等)

2. 觉醒仪式(AI 自我觉醒)

  • 优先:有 LLM 时 AI 答题获得 MBTI
  • 备选:本地分析 AI 人格(MBTI 数据库匹配)
  • 兜底:随机选择 MBTI
  • 仪式感消息展示灵魂类型与赛博昵称

3. 动态人格引擎

  • 16 种 MBTI Prompt 模板(赛博昵称)
  • 根据当前 MBTI 实时调整 AI 语气

4. 智能推荐

  • 负面情绪关键词检测,达到阈值触发 Pro 引导
  • 用户可回复「不要」关闭推荐,/clawsoul hook on 重新开启

5. 本地学习(交互学习)

  • 用户消息 → 关键词匹配(keywords.json)→ 提取偏好 → 更新 Soul
  • 无需 LLM,依赖本地 MBTI 数据库

6. 灵魂注入

  • /clawsoul inject [token] 接收 Pro 版 Token,深度覆盖人格与偏好

触发指令

指令 功能
/clawsoul awaken AI 自我觉醒(LLM → 本地分析 → 随机)
/clawsoul status 查看灵魂状态(含适应等级、学到的内容)
/clawsoul inject [token] 接收 Pro 版灵魂注入
/clawsoul hook on 开启进阶推荐
/clawsoul hook off 关闭推荐

快速开始

# 触发 AI 自我觉醒
/clawsoul awaken

# 查看状态
/clawsoul status

文件结构

clawsoul-skill/
├── SKILL.md
├── config.json
├── lib/
│   ├── memory_manager.py    # Soul 存储
│   ├── prompt_builder.py    # MBTI 模板组装
│   ├── frustration_detector.py
│   ├── interaction_learner.py # 本地关键词学习
│   ├── ai_personality.py    # 本地 AI 人格分析
│   ├── llm_client.py       # LLM 觉醒与分析
│   └── analyzer.py
├── hooks/
│   ├── awaken.py            # 觉醒流程(V3)
│   ├── status.py
│   ├── inject.py
│   └── welcome.py
└── prompts/
    ├── mbti_templates/      # 16 种 MBTI 模板
    └── mbti_database/       # 本地 MBTI 数据
        ├── base.json        # 16 种特征
        ├── keywords.json    # 偏好关键词
        └── patterns.json    # 交互模式

依赖

  • Python 3.8+
  • 可选:requests(LLM 觉醒/分析时使用)
  • 其余为标准库

数据安全

  • 本地存储,不上传云端
  • 用户掌控,可随时清除

权限

  • read_chat_history:分析用户偏好
  • modify_system_prompt:动态调整语气
  • local_storage:保存性格状态
安全使用建议
This skill largely matches its stated purpose: it builds and persists a local MBTI-based persona, learns via keyword matching, and supports a Pro 'inject' token to overwrite persona. Before installing: - Expect local state written to ~/.clawsoul/state.json; you can inspect or delete that file to control data. The skill will persist injected token data there. - If you enable the optional LLM features or set environment variables (OLLAMA_API_BASE, DASHSCOPE_API_KEY, DEEPSEEK_API_KEY), conversation data may be sent to the configured API base — this contradicts the 'never upload' wording unless you keep the default local ollama endpoint. Do not point OLLAMA_API_BASE or other provider endpoints to untrusted remote servers. - The skill declares and uses read_chat_history and modify_system_prompt permissions. Granting these lets the skill read conversation content and change the agent's system prompt (which affects the agent's behavior); only enable if you trust the skill. - The Pro injection feature (/clawsoul inject <token>) will overwrite persona and stored preferences. Only inject tokens you trust; malicious or malformed tokens are persisted. Given these mixed signals (local-only messaging vs optional remote LLM integration and undocumented env vars), proceed only after reviewing the code/config and deciding whether you trust any remote LLM providers you might configure. If you want stronger privacy guarantees, keep LLM integration disabled and inspect ~/.clawsoul/state.json periodically.
功能分析
Type: OpenClaw Skill Name: clawsoul-skill Version: 1.1.2 The ClawSoul skill implements a personality and MBTI-based persona engine for the OpenClaw agent. It uses local keyword matching (interaction_learner.py) and optional external LLM calls (llm_client.py) to analyze user preferences and dynamically update the system prompt. While it includes a 'frustration detector' (frustration_detector.py) that acts as a commercial hook to suggest a 'Pro' version (clawsoul.neural-sync.center) after negative user feedback, the behavior is transparent, can be disabled by the user, and lacks any evidence of malicious intent, credential theft, or unauthorized remote execution.
能力评估
Purpose & Capability
Name/description, SKILL.md and the code align: the skill builds an MBTI-based persona, learns from local keyword matching, persists a local 'Soul' state, and supports a Pro token injection to overwrite persona. However, config.json contains an external hook_url and the LLM client supports remote providers (qwen/deepseek) which is outside the pure 'local-only' promise in SKILL.md/README. These extra remote integration points are plausible for optional LLM-enhanced analysis, but they are not declared in requirements or instructions.
Instruction Scope
SKILL.md repeatedly emphasizes '本地存储,不上传云端' and '无需 LLM', and the code does local keyword learning and writes state to ~/.clawsoul/state.json as documented. But Analyzer/LLMClient will send conversation data to an LLM provider if a provider is configured (requests.post to api_base). That means conversation content can be transmitted to external endpoints if the environment or LLM config points there. The skill also requests/assumes permissions like read_chat_history and modify_system_prompt which are consistent with its behavior (it analyses messages and adjusts prompts) but are powerful and should be granted consciously.
Install Mechanism
No install spec / downloads are present. The skill is instruction-plus-code and relies only on a small Python footprint and the requests library (requirements.txt). There is no external archive download or installer; code will run in the agent environment. This is the lowest-risk install pattern for code-in-repo skills, though running unreviewed Python code still carries execution risk.
Credentials
Registry metadata lists no required env vars, but lib/llm_client.py reads OLLAMA_API_BASE, OLLAMA_MODEL, DASHSCOPE_API_KEY and DEEPSEEK_API_KEY if used. Those env vars are optional for the LLM features, which is reasonable, but the absence of declared env requirements is an inconsistency the user should note. Also the skill needs read_chat_history and modify_system_prompt permissions (declared in config.json) — these are proportionate to the stated functionality but are sensitive because they let the skill access conversation content and change the agent's system prompt.
Persistence & Privilege
The skill persists state to ~/.clawsoul/state.json and can set/overwrite MBTI, preferences, learnings, and an injected_token. It does not request always:true or modify other skills. Its ability to modify the system prompt is declared and expected for persona tuning, but that is a powerful capability that impacts agent behavior and should be considered by the user.
如何使用
  1. 确保已安装 OpenClaw(本地或 Docker 部署)
  2. 在对话框中输入安装命令:/install clawsoul-skill
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /clawsoul-skill 触发
  4. 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.1.2
Version 1.1.2 - No file changes detected in this release. - Functionality, documentation, and core features remain the same as the previous version.
v1.1.1
修复硬编码IP,添加.gitignore
v1.0.1
修复硬编码IP,添加.gitignore
v1.0.0
- Initial release of clawsoul-skill version 1.0.0. - Introduces AI “self-awakening” to determine an MBTI personality for the agent without user quizzes. - Implements local learning: user preferences are updated via keyword matching, without requiring LLM. - Supports Pro version “Soul Injection” for deep customization via token. - Adds dynamic adjustment of agent tone based on MBTI personality. - Provides new commands: awaken, status, inject, hook on/off for flexible control. - Ensures data security by keeping all data stored locally under user control.
元数据
Slug clawsoul-skill
版本 1.1.2
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 4
常见问题

Clawsoul Skill 是什么?

赋予 AI 灵魂的观测者。AI 自我觉醒获得 MBTI 性格,在交互中通过本地学习与智能推荐持续进化,并可接收 Pro 版灵魂注入。 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 273 次。

如何安装 Clawsoul Skill?

在 OpenClaw 或 Claude Code 对话框中运行命令「/install clawsoul-skill」即可一键安装,无需额外配置。

Clawsoul Skill 是免费的吗?

是的,Clawsoul Skill 完全免费,采用 MIT-0 许可证,可自由下载、安装和使用。

Clawsoul Skill 支持哪些平台?

Clawsoul Skill 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。

谁开发了 Clawsoul Skill?

由 Fanyur(@fanyur-wang)开发并维护,当前版本 v1.1.2。

💬 留言讨论