Skill 激活器
"你的 Skill 该醒了。"
Core Workflow
Three-layer process: Discover → Match → Fuse.
Layer 1: Discover Needs (需求发现)
Choose the appropriate discovery mode based on context:
Mode A — Environment Scan (数字生活体检)
- Run
bash scripts/scan_environment.sh to collect:
- Installed skills list with descriptions
- User identity from SOUL.md / USER.md / IDENTITY.md
- Connected channels (feishu, wecom, telegram, etc.)
- Workspace state (memory files, heartbeat config)
- Analyze scan results to identify:
- 🟢 Skills actively in use
- 🟡 Skills installed but underutilized (have the skill, not using it)
- 🔴 Capability gaps (needs not covered by any installed skill)
- Generate personalized activation report with specific, actionable recommendations
Mode B — Role-Based Recommendations (角色推荐)
- Identify user's role from SOUL.md or conversation
- Read
references/role-templates.md for role-specific pain points and automation ideas
- Cross-reference with installed skills
- Present top 3-5 recommendations ranked by: pain severity × ease of implementation
Mode C — Pain Point Interview (痛点挖掘对话)
- Ask user to walk through their daily workflow
- Listen for time-wasting patterns, repetitive tasks, context-switching pain
- Quantify: "You spend ~X minutes/day on Y — automatable?"
- Map each pain point to a concrete Skill combination
- Prioritize by time saved per week
Layer 2: Match Skills (智能匹配)
After discovering needs:
- Check installed skills that match the need
- Search ClawHub for additional skills:
clawhub search "<need description>"
- Identify gaps — needs no existing skill covers
- Present matching plan:
- ✅ Already have: installed skills that apply
- 📦 Recommend install: available on ClawHub via
clawhub install <name>
- 🔧 Can fuse: combine existing skills to cover this need
- 🆕 Need to create: no skill exists, suggest building one
Layer 3: Fuse & Generate (融合生成)
When combining skills into a new workflow:
- Read
references/fusion-guide.md for fusion patterns and templates
- Read each source skill's SKILL.md to understand inputs, outputs, dependencies
- Design the pipeline (sequential, fan-out, aggregation, or enhancement)
- Generate new fused SKILL.md with proper frontmatter, workflow steps, and error handling
- Add any glue scripts to
scripts/ if data format conversion is needed
- Package with
package_skill.py for distribution
Output Format
Activation Report
🔍 Skill 激活报告
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
👤 身份：{role from SOUL.md}
🔌 已连接：{channels}
📦 已安装：{N} 个 Skill
💡 发现 {N} 个自动化机会：
1️⃣ {pain point description}
⏱ 预计每周节省：{X} 分钟
🧩 需要：{Skill A} + {Skill B}
📊 可行度：⭐⭐⭐⭐⭐
→ [一键激活]
2️⃣ {pain point description}
...
🟡 沉睡的 Skill（装了没用）：
- {skill name}: 它能帮你 {what it does}
📦 推荐安装：
- {skill name}: 解决 {what problem}
- 安装命令：clawhub install {name}
Fused Skill Output
When generating a fused skill, output:
- The complete SKILL.md content
- Any required glue scripts
- Installation/usage instructions
- Packaging command:
package_skill.py <path>
Key Principles
- Be specific, not generic: "你每周花2小时做周报" > "你可以自动化一些任务"
- Quantify savings: Always estimate time/effort saved
- One-click actionable: Every recommendation should be immediately executable
- Respect what's installed: Prioritize solutions using skills the user already has
- Progressive: Start with quick wins, then suggest more advanced automations
