Skill Combo Recommender

Recommend optimal skill combinations and workflows for user tasks, offering usage examples and preset workflows to optimize project execution.

Security Scan

The skill's code, CLI, and instructions are coherent with a recommender for skill combinations and workflows; nothing requests unrelated credentials or installs code from untrusted URLs, though the package includes a local installer script that you should review before running.

Purpose & Capability

The name/description match the included Python implementation: a local SKILLS database, WORKFLOWS, keyword mappings, and CLI behaviors for recommending skill combinations. The listed example recommendations (content, video, data workflows) align with the skill's stated purpose. The presence of skills like a proxy manager is part of the catalog and makes sense for a recommender that lists third‑party skills, even if some of those recommended skills have potentially sensitive functionality themselves.

Instruction Scope

SKILL.md contains only recommendation logic, CLI usage examples, and instructions for extending the local skill/workflow databases. It does not instruct the agent to read unrelated system files or to exfiltrate data. It does mention possible future features (e.g., "record user choices" for learning), but that is only a roadmap item, not an instruction that will run now.

Install Mechanism

There is an install.sh that copies files into the user's ~/.openclaw/skills and optionally creates a symlink in ~/.local/bin; this is a normal user-space install. Note: the repository metadata summary earlier said "instruction-only", which is inconsistent with the presence of install.sh and a declared entrypoint in skill.json. The install script performs only local file operations and no network downloads.

Credentials

The skill requests no environment variables or credentials and only requires python3 (declared in skill.json). That is proportional to a local CLI recommender. No secrets or unrelated credentials are requested.

Persistence & Privilege

The skill does not set always:true and does not request elevated system privileges. The installer writes to user-owned directories (~/.openclaw/skills and optionally ~/.local/bin), which is standard for user-space tools.

Assessment

This skill appears coherent and implements a local recommender. Before installing: (1) review install.sh to ensure you are comfortable with it copying files to ~/.openclaw/skills and creating a symlink in ~/.local/bin; (2) scan the full source (source/skill_combo_recommender.py) for any network calls, telemetry, or subprocess calls if you want extra assurance—current excerpts show only local data structures; (3) be aware that the recommender lists other skills (e.g., proxy managers) which themselves may perform sensitive actions—install and run those only if you trust them; and (4) if you see claims about "learning user behavior" or telemetry in a later version, verify where that data is stored/sent before enabling such features.

智能推荐技能组合，帮助用户找到最佳工作流程。

核心功能

1. 任务分析：根据用户描述的任务，分析需要的技能
2. 技能组合推荐：推荐最佳技能组合和工作流程
3. 使用示例：提供具体的使用示例和命令
4. 预设工作流：提供常见场景的预设工作流（小红书运营、内容创作、视频制作等）

使用场景

场景 1：小红书运营 用户想要运营小红书账号

• 推荐技能：xiaohongshu-content + xiaohongshu-image-gen + xiaohongshu-proxy-manager + social-media-scheduler
• 工作流程：内容创作 → 图片生成 → 发布排期
• 使用示例：

  # 1. 生成爆款内容
  xiaohongshu-content --category 家装 --style 种草
  
  # 2. 生成配图
  xiaohongshu-image-gen --prompt 现代简约风格客厅设计
  
  # 3. 设置排期发布
  social-media-scheduler --platform xiaohongshu --schedule "2026-03-20 10:00"
  

场景 2：内容创作流水线 用户想要高效创作多平台内容

• 推荐技能：content-researcher + social-publisher + ai-content-tailor + wechat-formatter
• 工作流程：内容调研 → 内容裁剪 → 格式化 → 多平台发布
• 使用示例：

  # 1. 调研热门话题
  content-researcher --topic 人工智能 --days 7
  
  # 2. 裁剪为多平台版本
  ai-content-tailor --input article.md --platforms xiaohongshu,zhihu,wechat
  
  # 3. 发布到多平台
  social-publisher --input xiaohongshu.md --platform xiaohongshu
  

场景 3：视频内容制作 用户想要制作视频内容

• 推荐技能：video-frames + auto-subtitle + video-generate + text-to-podcast
• 工作流程：视频生成 → 提取帧 → 生成字幕 → 音频配音
• 使用示例：

  # 1. 生成视频
  video-generate --prompt AI助手介绍 --duration 15s
  
  # 2. 生成字幕
  auto-subtitle --input video.mp4 --language zh-CN
  
  # 3. 文本转音频
  text-to-podcast --input script.txt --voice male
  

场景 4：数据分析和可视化 用户想要分析数据并生成图表

• 推荐技能：stock-analyzer + data-chart-tool + tushare-finance
• 工作流程：数据获取 → 数据分析 → 图表生成
• 使用示例：

  # 1. 获取股票数据
  tushare-finance --code 000001 --period daily
  
  # 2. 分析股票
  stock-analyzer --symbol 000001 --indicators MA,RSI,MACD
  
  # 3. 生成图表
  data-chart-tool --input data.csv --type line --title 股票走势
  

核心逻辑

技能标签系统

每个技能都有标签，用于匹配用户需求：

内容创作：xiaohongshu-content, content-researcher, social-publisher, ai-content-tailor, wechat-formatter 视频制作：video-generate, video-frames, auto-subtitle, text-to-podcast 数据分析：stock-analyzer, data-chart-tool, tushare-finance 文件管理：batch-renamer, photo-organizer, download-organizer, video-organizer, music-tagger, file-sorter 音频处理：audio-note-taker, auto-subtitle, text-to-podcast 项目管理：skill-composer, social-media-scheduler, email-ai-assistant AI 工具：summarize, openai-image-gen, openai-whisper

推荐算法

1. 关键词匹配：根据用户描述的关键词匹配技能标签
2. 权重排序：根据匹配度排序推荐技能
3. 工作流生成：将推荐技能组织成工作流程
4. 示例生成：为每个推荐技能提供使用示例

预设工作流

小红书运营全流程：

• 技能：xiaohongshu-content + xiaohongshu-image-gen + xiaohongshu-proxy-manager + social-media-scheduler
• 步骤：1. 生成爆款内容 → 2. 生成配图 → 3. 设置代理 → 4. 排期发布

内容创作流水线：

• 技能：content-researcher + social-publisher + ai-content-tailor + wechat-formatter
• 步骤：1. 调研热门话题 → 2. 裁剪内容 → 3. 格式化 → 4. 多平台发布

视频内容制作：

• 技能：video-generate + video-frames + auto-subtitle + text-to-podcast
• 步骤：1. 生成视频 → 2. 提取帧 → 3. 生成字幕 → 4. 音频配音

数据分析和可视化：

• 技能：stock-analyzer + data-chart-tool + tushare-finance
• 步骤：1. 获取数据 → 2. 分析数据 → 3. 生成图表

CLI 命令

# 基础用法：根据任务描述推荐技能
skill-combo-recommender --task "我想运营小红书账号"

# 指定平台：推荐特定平台的技能组合
skill-combo-recommender --task "内容创作" --platform xiaohongshu,wechat

# 查看预设工作流：列出所有预设工作流
skill-combo-recommender --list-workflows

# 使用预设工作流：使用特定的工作流
skill-combo-recommender --workflow "小红书运营全流程"

# 查看技能库：列出所有可用技能和标签
skill-combo-recommender --list-skills

# 生成文档：生成推荐的工作流文档
skill-combo-recommender --task "内容创作" --output markdown

扩展性

添加新技能

在 source/skill_database.py 中添加新技能：

skills = {
    "skill-name": {
        "name": "技能名称",
        "description": "技能描述",
        "tags": ["标签1", "标签2"],
        "examples": ["示例1", "示例2"]
    },
    # ... 添加更多技能
}

添加新工作流

在 source/workflow_database.py 中添加新工作流：

workflows = {
    "workflow-name": {
        "name": "工作流名称",
        "description": "工作流描述",
        "skills": ["skill1", "skill2", "skill3"],
        "steps": ["步骤1", "步骤2", "步骤3"]
    },
    # ... 添加更多工作流
}

优化方向

1. 学习用户行为：记录用户选择的工作流，优化推荐算法
2. 个性化推荐：根据用户历史使用记录，提供个性化推荐
3. 工作流模板：提供更多预设工作流模板
4. 集成 skill-composer：直接生成 skill-composer 配置文件