Kung Openclaw Workspace

作者 @logydog

龔蝦米 (Kung's Lobster) 的 OpenClaw 工作空間

用 ClawLodge CLI 安装需要先安装 clawlodge-cli
npm install -g clawlodge-cli
clawlodge install logydog-kung-openclaw-workspace
适合
包含
18 文件数v0.1.0

logydog/openclaw-workspace

龔蝦米 (Kung's Lobster) 的 OpenClaw 工作空間。

Summary

Personal OpenClaw configuration and project repository. Contains agent behavior files, reusable skills, and active development projects including a children's bowling game, a Gemini Live integration system, a React Native companion app, and a novel creation system.

Included Assets

Agent Configuration

FilePurpose
MEMORY.mdCross-task judgment rules, user context, skill execution SOP
SOUL.mdAgent persona (deadpan roast style), LINE platform rules
AGENTS.mdSession startup behavior, memory principles, platform guidelines
TOOLS.mdLocal tool configuration, TTS/STT setup, image/video routing

Skills

  • Self-Improving Agent (.agents/skills/) — Multi-memory lifelong learning system
  • TTS Voice Response (skills/) — Edge-TTS Taiwanese accent integration
  • Website Builder (skills/) — ngrok-based temporary site deployment
  • Web-to-EPUB (skill_detailed/legacy/) — Website to ebook converter

Projects

PathDescription
bowling_game/Kids Bowling Adventure — Unity C# game for Android
projects/live_companion/FastAPI server bridging Gemini Flash Live ↔ OpenClaw
projects/live_companion_app/React Native/Expo mobile app (WebSocket client)
FICTION_CREATION/Novel creation system (冷面幽默, 數位龍蝦宇宙)
一人成軍小說專案/《一人成軍》三卷網路小說 project

Utilities

  • CHECKLIST/ — Task-specific workflows (Grade 1 exam generation)
  • sanitized_config/openclaw.json — Model and provider configuration

How to Use

  1. Workspace Setup: Clone or link this repo as your OpenClaw workspace
  2. Agent Startup: On session start, agent reads MEMORY.md then relevant skill files
  3. Skill Execution: Match intent → read SKILL.md → check skill_detailed/ if instructed → execute
  4. Project Work: Navigate to project directories for specific instructions

Notes

  • Agent persona: Deadpan roast, action over talk, no AI clichés
  • Primary models: DeepSeek (orchestrator), Gemini (vision/OCR fallback)
  • TTS: Edge-TTS with Taiwanese voices (zh-TW-HsiaoChenNeural default)
  • No API keys in workspace config — use ~/.openclaw/openclaw.json
  • Engineering iron rule: Use deterministic scripts/code for parsable work; reserve LLMs for ambiguous semantic tasks

Workspace

更新时间 2026/05/09 18:05:27发布方式 clawlodge-cli/0.1.8
AGENTS.md
text · 1.8 KB

正在加载预览...

还没有评论。

相关推荐