给 Claude Code 加了个 skill，强迫它先想清楚再动手
看完 Trae 的 MTC 模式，我在 Claude Code 里做了个 /mtc skill。6 个阶段，不许跳步，先想清楚再让 AI 写代码。做游戏的时候效果比我预期好。
- AI (218)
- Game Design (6)
- Claude Code (39)
- AI Agents (27)
- 多 Agent (6)
- 产品设计 (4)
- 游戏 (1)
- 开源 (4)
- 架构决策 (1)
- AgentScope (1)
- 架构设计 (1)
- Agora (3)
- LLM (12)
- 狼人杀 (2)
- 桌面宠物 (1)
- Clawd (10)
- Desktop Pet (1)
- Product Design (2)
- Prompt Engineering (6)
- 人格系统 (1)
- Personality System (1)
- 记忆系统 (1)
- Memory Systems (1)
- 架构 (1)
- Electron (2)
- Architecture (1)
- Open Source (1)
- AI Workflow (2)
- Skills (2)
- Organization (8)
- Future of Work (8)
- 投资 (10)
- 思维模型 (10)
- Investing (10)
- Mental Models (10)
- Agent Fleet (6)
- AI Companion (2)
- SaaS (2)
- AI Safety (2)
- Interpretability (2)
- Anthropic (2)
- MCP (2)
- Knowledge Management (2)
- Obsidian (2)
- Agents (110)
- Foundry (2)
- Easter Eggs (2)
- Fun (2)
- Software Development (24)
- Frontend (4)
- Agent Runtimes (3)
- 自动化 (2)
- macOS (4)
- Automation (2)
- Compute (4)
- Philosophy (2)
- 微信 (1)
- WeChat (1)
- Chips (2)
- Hardware (2)
- GPU (2)
- TPU (2)
- Mio (66)
- Lumi (8)
- Writing (8)
- NLP (6)
- Product (20)
- Engineering (18)
- Chinese (2)
- Design (8)
- Voice (8)
- DevOps (8)
- Linux (8)
- VPN (4)
- 翻墙 (2)
- 科学上网 (2)
- Self-hosting (2)
- Career (4)
- Economics (4)
- China (3)
- Psychology (3)
- Labor (2)
- Memory (4)
- Claude (4)
- GPT (2)
- Business (2)
- OpenClaw (8)
- TTS (2)
- GCP (4)
- Workflow (2)
- Ops (2)
- Cost Optimization (2)
- Agent Teams (4)
- PanPanMao (4)
- Best Practices (1)