skill-generator

name	skill-generator
description	Meta-skill that generates new skills based on user requirements. Enables self-evolution.

Skill Generator (Self-Evolution System)

🚀 Revolutionary Feature

This is a meta-skill that allows the AI agent to evolve itself by creating new skills when needed!

When to use this skill

Use when:

• User requests functionality not covered by existing skills
• New blockchain features need to be supported
• Custom analysis or automation is needed
• Extending the agent's capabilities

How It Works

Phase 1: Needs Analysis

User: "我想查询NFT的交易记录"
Agent: 🤔 Analyzing request...
       Current skills: ❌ No NFT skill found
       → Triggering Skill Generator

Phase 2: Planning Generation

Agent generates planning:

📋 New Skill Plan: "nft-transaction-tracker"
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Purpose: Track NFT transactions and ownership history

Key Features:
  1. Query NFT transfer events
  2. Get NFT ownership history
  3. Calculate NFT floor price
  4. Detect wash trading

Data Sources:
  - TronScan NFT API
  - Marketplace APIs (ApeNFT, etc.)

Implementation:
  - SKILL.md: Documentation
  - scripts/track_nft.py: Core logic
  - Integration: MCP tool wrapper

Estimated Complexity: Medium (3-4 hours)

❓ Approve this plan? (yes/no)

Phase 3: Implementation (After Approval)

User: yes

Agent: ✅ Generating skill...
       [====================] 100%
       
       Created files:
       ✓ skills/nft-transaction-tracker/SKILL.md
       ✓ skills/nft-transaction-tracker/scripts/track_nft.py
       ✓ Updated src/tool_wrappers.py
       ✓ Registered in src/main.py
       
       🎉 New skill ready!

Phase 4: Verification & Save

Agent: 🧪 Testing new skill...
       ✓ Basic functionality works
       ✓ API connections successful
       
       💾 Save this skill permanently? (yes/no)

User: yes

Agent: ✅ Skill saved and activated!
       You can now use: get_nft_transactions()

Workflow

graph TD
    A[User Request] --> B{Existing Skills?}
    B -->|Yes| C[Use Existing]
    B -->|No| D[Generate Planning]
    D --> E[User Review]
    E -->|Reject| F[Refine Plan]
    E -->|Approve| G[Generate Code]
    F --> D
    G --> H[Test Skill]
    H --> I{Works?}
    I -->|No| J[Debug & Fix]
    I -->|Yes| K[User Review Code]
    J --> H
    K -->|Save| L[Activate Skill]
    K -->|Discard| M[End]
    L --> N[Available Forever]

Generated Skill Structure

Each generated skill follows standard format:

skills/[skill-name]/
├── SKILL.md              # Auto-generated docs
├── scripts/
│   └── main.py          # Core implementation
└── examples/            # Usage examples (optional)

Safety & Quality

✅ Built-in Safeguards

1. Code Review: User must approve before execution
2. Sandboxed Testing: New skills tested in isolation
3. Rollback: Can delete if not working
4. Version Control: Track all generated skills

🔍 Quality Checks

• ✓ Follows existing skill conventions
• ✓ Includes error handling
• ✓ API usage is safe
• ✓ Documentation is clear

Example Use Cases

1. NFT Tracking

User: "跟踪我的NFT交易"
→ Generates: nft-transaction-tracker

2. DeFi Analytics

User: "分析流动性池的收益率"
→ Generates: liquidity-pool-analyzer

3. Custom Alerts

User: "当TRX价格跌破0.1时通知我"
→ Generates: price-alert-monitor

4. Batch Operations

User: "批量给100个地址转账"
→ Generates: batch-transfer-executor

Limitations

Cannot Generate:

• ❌ Skills requiring private APIs without credentials
• ❌ Skills violating TRON network rules
• ❌ Malicious or harmful functionality
• ❌ Skills requiring external dependencies not in project

Can Generate:

• ✅ Any read-only blockchain queries
• ✅ Transaction builders (unsigned)
• ✅ Data analysis and reporting
• ✅ Automation workflows

Technical Details

Planning Template

Uses structured planning:

• Purpose: What problem does it solve?
• Features: What can it do?
• Data Sources: Which APIs?
• Implementation: Technical approach
• Testing: How to verify?

Code Generation

Uses LLM to generate:

1. SKILL.md documentation
2. Python implementation
3. Error handling
4. MCP integration code

Auto-Registration

Automatically updates:

• src/tool_wrappers.py - Adds wrapper function
• src/main.py - Registers MCP tool

Future Enhancements

• 🔄 Skill versioning and updates
• 📊 Usage analytics (which skills are most used)
• 🤝 Skill combination (chain multiple skills)
• 🌐 Community skill sharing
• 🧠 Learning from user feedback

Meta Philosophy

"The best agent is one that can teach itself new tricks."

This skill generator embodies the principle of continuous evolution. Instead of being limited by pre-programmed capabilities, the agent can adapt to any blockchain task thrown at it!