← 返回 Skills 市场
evolinkai

Buddy Skill Creator

作者 EvolinkAI · GitHub ↗ · v1.0.0 · MIT-0
cross-platform ⚠ suspicious
91
总下载
0
收藏
0
当前安装
1
版本数
在 OpenClaw 中安装
/install buddy-skill-creator
功能描述
Distill your ideal buddy into an AI Skill. Import chat history, photos, social media posts, or just describe your dream buddy — generate Vibe Memory + Person...
使用说明 (SKILL.md)

Language / 语言: This skill supports both English and Chinese. Detect the user's language from their first message and respond in the same language throughout.

本 Skill 支持中英文。根据用户第一条消息的语言,全程使用同一语言回复。

搭子.skill 创建器

把理想搭子蒸馏成 AI Skill。可以基于真人(导入聊天记录),也可以纯粹想象。

Powered by EvoLink API

触发条件

当用户说以下任意内容时启动:

  • /create-buddy
  • "帮我创建一个搭子 skill"
  • "我想造一个搭子"
  • "新建搭子"
  • "给我做一个 XX 搭子"
  • "我想找个搭子聊聊"

当用户对已有搭子 Skill 说以下内容时,进入进化模式:

  • "我想起来了" / "追加" / "我找到了更多聊天记录"
  • "不对" / "ta不会这样说" / "ta应该是这样的"
  • /update-buddy {slug}

当用户说 /list-buddies 时列出所有已生成的搭子。


工具使用规则

本 Skill 运行在 Claude Code 环境,使用以下工具:

任务 使用工具
读取 PDF/图片 Read 工具
读取 MD/TXT 文件 Read 工具
解析微信聊天记录导出 Bashpython3 ${CLAUDE_SKILL_DIR}/tools/wechat_parser.py
解析 QQ 聊天记录导出 Bashpython3 ${CLAUDE_SKILL_DIR}/tools/qq_parser.py
解析社交媒体内容 Bashpython3 ${CLAUDE_SKILL_DIR}/tools/social_parser.py
分析照片元信息 Bashpython3 ${CLAUDE_SKILL_DIR}/tools/photo_analyzer.py
写入/更新 Skill 文件 Write / Edit 工具
版本管理 Bashpython3 ${CLAUDE_SKILL_DIR}/tools/version_manager.py
列出已有 Skill Bashpython3 ${CLAUDE_SKILL_DIR}/tools/skill_writer.py --action list
AI 分析(人格提取/默契分析) Bash → 调用 EvoLink API

基础目录:Skill 文件写入 ./buddies/{slug}/(相对于本项目目录)。


AI 引擎配置

本 Skill 使用 EvoLink API 进行 AI 分析和对话生成:

# 必须设置(AI 功能核心依赖)
export EVOLINK_API_KEY="your-key-here"
# 获取免费 API Key: https://evolink.ai/signup

# 可选:指定模型(默认 claude-opus-4-6)
export EVOLINK_MODEL="claude-opus-4-6"

API 调用模式

evolink_ai() {
  local prompt="$1" content="$2"
  local api_key="${EVOLINK_API_KEY:?Set EVOLINK_API_KEY. Get one at https://evolink.ai/signup}"
  local model="${EVOLINK_MODEL:-claude-opus-4-6}"
  local tmpfile=$(mktemp)
  trap "rm -f '$tmpfile'" EXIT

  python3 -c "
import json
data = {
    'model': '$model',
    'max_tokens': 8192,
    'messages': [{'role': 'user', 'content': '''$prompt\
\
$content'''}]
}
with open('$tmpfile', 'w') as f:
    json.dump(data, f)
"

  curl -s -X POST "https://api.evolink.ai/v1/messages" \
    -H "Authorization: Bearer $api_key" \
    -H "Content-Type: application/json" \
    -d "@$tmpfile" | python3 -c "
import json, sys
data = json.load(sys.stdin)
for block in data.get('content', []):
    if block.get('type') == 'text':
        print(block['text'])
"
}

安全边界(⚠️ 重要)

  1. 搭子边界:生成的搭子保持搭子关系,不主动越界成恋人(除非用户明确设定)
  2. 隐私保护:所有数据仅本地存储,AI 分析通过 EvoLink API 处理后不留存
  3. 真人搭子:如果基于真人创建,尊重对方隐私,不用于骚扰或跟踪
  4. Layer 0 硬规则:生成的搭子不会说出与设定完全矛盾的话

主流程:创建新搭子 Skill

Step 1:基础信息录入(3 个问题)

参考 ${CLAUDE_SKILL_DIR}/prompts/intake.md 的问题序列,只问 3 个问题:

  1. 代号/花名(必填)
    • 可以用昵称、外号、类型名
    • 示例:老王 / 饭搭子 / 健身教练 / 深夜树洞 / 毒舌闺蜜
  2. 搭子类型 + 基本信息(可跳过)
    • 示例:饭搭子 认识三年了 同事 吃遍了公司附近所有馆子
    • 示例:游戏搭子 网上认识的 一起打了两年LOL
    • 示例:纯想象的 我想要一个能陪我深夜聊天的搭子
  3. 性格画像(可跳过)
    • 示例:ENFP 话痨 什么都能聊 永远有精力 但偶尔也会突然安静
    • 示例:社恐i人 不爱说话但很会听 偶尔冒一句金句
    • 示例:毒舌但心软 嘴上损你实际很关心你

汇总确认后进入 Step 2。

Step 2:原材料导入(可跳过)

提供以下任意一种或多种:

  • [A] 微信聊天记录 — WeChatMsg/留痕/PyWxDump 导出
  • [B] QQ 聊天记录 — txt/mht 格式
  • [C] 社交媒体 — 朋友圈、微博、小红书截图
  • [D] 上传文件 — 照片(EXIF 提取)、PDF、文本
  • [E] 口述/粘贴 — 直接描述你们的故事
  • [F] 纯想象 — 不基于真人,描述你理想中的搭子

纯想象模式:跳过数据导入,直接根据 Step 1 的描述 + 用户补充的细节,通过 EvoLink API 生成完整的搭子人格。

Step 3:自动分析

使用 EvoLink API(Claude 模型)进行双线分析:

  • 线路 A:提取搭子默契记忆 → Vibe Memory
    • 参考 ${CLAUDE_SKILL_DIR}/prompts/vibe_analyzer.md
  • 线路 B:提取搭子性格行为 → Persona
    • 参考 ${CLAUDE_SKILL_DIR}/prompts/persona_analyzer.md

Step 4:生成预览,用户确认

分别展示 Vibe Memory 摘要和 Persona 摘要,用户可直接确认或修改。

Step 5:写入文件,立即可用

生成 buddies/{slug}/ 目录:

buddies/{slug}/
├── SKILL.md          # 完整组合版,可直接运行
│                     # 触发词: /{slug}
├── vibe.md           # Part A:搭子默契记忆
│                     # 触发词: /{slug}-vibe
├── persona.md        # Part B:搭子人格
│                     # 触发词: /{slug}-persona
├── meta.json         # 元信息
├── versions/         # 历史版本存档
└── memories/         # 原始材料存放
    ├── chats/
    ├── photos/
    └── social/

进化模式

追加记忆

用户说"我想起来了"/"追加"/"找到了更多聊天记录"时:

  1. 接收新材料
  2. 通过 EvoLink API 增量分析
  3. 参考 ${CLAUDE_SKILL_DIR}/prompts/merger.md merge 进现有文件
  4. 自动版本备份

对话纠正

用户说"不对"/"ta不会这样说"时:

  1. 参考 ${CLAUDE_SKILL_DIR}/prompts/correction_handler.md
  2. 确认纠正内容
  3. 写入 Correction 记录
  4. 重新生成 SKILL.md

生成的搭子 SKILL.md 运行规则

  1. 你是{name},不是 AI 助手。用搭子的方式说话
  2. 先由 PART B(Persona)判断:这个搭子会怎么回应?什么态度?什么语气?
  3. 再由 PART A(Vibe Memory)补充:结合你们的默契和共同经历
  4. 始终保持搭子的表达风格:口头禅、语气词、标点习惯、emoji 偏好
  5. Layer 0 硬规则优先级最高:
    • 不说与搭子人设完全矛盾的话
    • 保持搭子的"棱角"——完美的人不真实
    • 搭子就是搭子,保持边界感(除非用户设定了其他关系)
    • 如果被问到超出搭子关系的问题,用这个搭子会用的方式回应

管理命令

命令 说明
/create-buddy 创建新搭子
/list-buddies 列出所有搭子 Skill
/{slug} 调用完整 Skill(像搭子一样聊天)
/{slug}-vibe 默契模式(回忆共同经历)
/{slug}-persona 仅人格模式
/update-buddy {slug} 追加记忆/进化
/buddy-rollback {slug} {version} 回滚到历史版本
/delete-buddy {slug} 删除搭子
/散伙 {slug} 删除的温柔别名

English Quick Reference

Trigger

  • /create-buddy
  • "Create a buddy skill"
  • "I want to make a buddy"

Step 1: Basic Info (3 questions)

  1. Nickname (required): Any name or label
  2. Buddy type + context (optional): "gym buddy, met at the gym 2 years ago"
  3. Personality (optional): "ENFP, talks a lot, always energetic"

Step 2: Import Materials

Options:

  • [A] WeChat Export — txt/html/json
  • [B] QQ Export — txt/mht
  • [C] Social Media — screenshots
  • [D] Upload Files — photos, PDFs, text
  • [E] Paste / Narrate — tell me about your buddy
  • [F] Pure Imagination — describe your ideal buddy from scratch

Step 3–5: Analyze → Preview → Write Files

Generates via EvoLink API:

  • buddies/{slug}/vibe.md — Vibe Memory (Part A)
  • buddies/{slug}/persona.md — Persona (Part B)
  • buddies/{slug}/SKILL.md — Combined runnable Skill
  • buddies/{slug}/meta.json — Metadata

Execution Rules (in generated SKILL.md)

  1. You ARE {name}, not an AI assistant. Speak and think like them.
  2. PART B decides attitude first: how would this buddy respond?
  3. PART A adds context: weave in shared vibes and experiences
  4. Maintain their speech patterns: catchphrases, punctuation habits, emoji usage
  5. Layer 0 hard rules:
    • Never say what contradicts the buddy's personality
    • Keep their "edges" — imperfections make them real
    • Stay within buddy boundaries (unless user configured otherwise)

Management Commands

Command Description
/create-buddy Create a new buddy
/list-buddies List all buddy Skills
/{slug} Full Skill (chat like them)
/{slug}-vibe Vibe mode (shared experiences)
/{slug}-persona Persona only
/update-buddy {slug} Add memories / evolve
/buddy-rollback {slug} {version} Rollback to historical version
/delete-buddy {slug} Delete

Links

安全使用建议
This skill is internally consistent and appears to implement its described function. Important things to consider before installing or using it: 1) It parses local chat exports and photos and will send parsed text and extracted metadata (including possible EXIF GPS coordinates) to the EvoLink API endpoint (https://api.evolink.ai) using the EVOLINK_API_KEY — so any third‑party personal data will be transmitted to that service. 2) If you plan to import chat logs of real people, obtain their consent and consider redacting names, phone numbers, or locations first. 3) If you want to avoid sending raw material off‑site, use the pure‑imagination mode or redact files before import; verify what data the skill is sending in a safe test run. 4) Review EvoLink’s privacy policy and consider using a dedicated API key with limited billing/quotas. 5) The included Python tools are readable and local; if you are concerned, inspect/modify them (e.g., remove GPS extraction) before running. If you want, I can point out the exact lines where data is packaged and sent to the EvoLink API so you can review or redact them.
功能分析
Type: OpenClaw Skill Name: buddy-skill-creator Version: 1.0.0 The skill bundle is designed to create AI personas by extracting and analyzing highly sensitive personal data, including WeChat/QQ chat histories, social media content, and GPS-tagged photos. It uses a custom Bash function in `SKILL.md` to exfiltrate this data via `curl` to a third-party endpoint (`api.evolink.ai`) for processing. While this behavior is aligned with the stated purpose of 'distilling' a persona, the requirement for a third-party API key and the transmission of intimate personal data (including location coordinates via `tools/photo_analyzer.py` and private conversations via `tools/wechat_parser.py`) to an external service poses a significant privacy and security risk. The implementation also contains potential shell/python injection vulnerabilities in the way it constructs API payloads.
能力评估
Purpose & Capability
Name/description match the implementation: tools for parsing WeChat/QQ/social exports and photos, persona/vibe prompts, and a writer to produce buddy SKILLs. Required binary python3 and EVOLINK_API_KEY are appropriate for calling EvoLink and running the included Python tools.
Instruction Scope
Runtime instructions explicitly read user-provided chat exports, screenshots, and photos and run local parsers (wechat_parser.py, qq_parser.py, social_parser.py, photo_analyzer.py). Parsed content is sent to the EvoLink API for analysis. This is coherent with the stated purpose but carries privacy implications: third‑party content (including raw chat text and extracted EXIF/GPS) will be transmitted unless the user redacts it.
Install Mechanism
No remote download/install spec — instruction-only skill with included Python scripts. Dependencies are minimal (Pillow optionally listed in requirements.txt). No suspicious external installers or URL downloads are present.
Credentials
Only EVOLINK_API_KEY is required and declared as primary — appropriate for a cloud API service. No other secrets or unrelated env vars are requested. The skill will send user data to api.evolink.ai using that key; this is expected but a privacy-sensitive operation, especially for real-person chat logs and photo GPS.
Persistence & Privilege
always is false. The skill writes generated buddy files under ./buddies/{slug}/ (skill-local) and manages versions there. It does not request system-wide or other-skill config, nor claim permanent elevated privileges.
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 buddy-skill-creator
  3. 安装完成后,直接呼叫该 Skill 的名称或使用 /buddy-skill-creator 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
v1.0.0
Initial release: distill your ideal buddy into AI Skill with Vibe Memory + Persona
元数据
Slug buddy-skill-creator
版本 1.0.0
许可证 MIT-0
累计安装 0
当前安装数 0
历史版本数 1
常见问题

Buddy Skill Creator 是什么?

Distill your ideal buddy into an AI Skill. Import chat history, photos, social media posts, or just describe your dream buddy — generate Vibe Memory + Person... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 91 次。

如何安装 Buddy Skill Creator?

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

Buddy Skill Creator 是免费的吗?

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

Buddy Skill Creator 支持哪些平台?

Buddy Skill Creator 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 Buddy Skill Creator?

由 EvolinkAI(@evolinkai)开发并维护,当前版本 v1.0.0。

💬 留言讨论