A ready-to-use AI Agent workspace template: built on a GitHub repository, integrating long-term memory, reusable AI skill packs, GitHub Actions automation, and the Copilot coding agent — making AI your long-term collaborative partner.
github-claw demonstrates how to turn a GitHub repository into an AI agent with persistent memory, reusable skills, and full automation. The core is not a specific product — it is a mechanism that keeps AI working continuously:
- 🧠 Files as Memory: Preserve AI Agent context across conversations using repository files, achieving true long-term memory beyond the single-session limit.
- 🔧 AI Skill Packs (Agent Skills): Encapsulate capabilities like UI/UX design, AI image generation, and SEO auditing into standardized skill packs that the AI assistant can invoke directly and anyone can reuse.
- ⚡ GitHub Actions-Driven AI Automation: Use GitHub Actions to run a daily tracker of new Pull Requests in the CesiumGS/cesium project, handle Issue triage, auto-assign the Copilot coding agent — all without human intervention.
- 🚀 Vibe Coding in Practice: Describe requirements in natural language; the AI Agent handles coding, testing, and deployment automatically.
When starting a new Copilot conversation, tell the assistant:
Please read AGENTS.md and MEMORY.md, restore your identity and work state, then continue our work.
This lets the AI assistant Claw restore its identity, load long-term memory, and resume from where things left off.
A Skill Pack (Agent Skill) is the core mechanism of this repository — it encapsulates reusable AI capabilities as standard units stored under .agents/skills/<skill-name>/, with SKILL.md as the main entry point. The AI assistant can invoke any skill directly when given a task.
Built-in skill packs:
| Skill Pack | Description |
|---|---|
ui-ux-pro-max |
UI/UX design system: 50+ visual styles, 161 color schemes, 57 font pairings — generate high-quality interfaces instantly |
ai-image-generation |
Multi-model AI image generation (FLUX, Gemini, Grok, etc.) — text-to-image, image-to-image, LoRA, and more |
seo-audit |
SEO audit and optimization suggestions — auto-analyze page structure, keyword density, and technical SEO issues |
Skill workflow:
- Receive a task → check local
.agents/skills/first, reuse existing skill packs when available - No suitable local skill → search GitHub open-source repos or Skills.sh
- After installing, store under
.agents/skills/<skill-name>/and update the index
| Level | File Location | Content |
|---|---|---|
| Long-term memory | MEMORY.md |
User preferences, project goals, persistent agreements |
| Daily log | memory/YYYY-MM-DD.md |
AI work records, decisions, and notes for the day |
| Task tracking | memory/tasks.md |
Cross-session to-dos and progress |
| Skill assets | .agents/skills/ |
Installed, callable AI Agent skill packs |
| Workflow | Trigger | AI Capability |
|---|---|---|
cesium-daily-pr.yml |
Daily at 13:00 | Automatically fetches new PRs opened in CesiumGS/cesium that day and publishes a summary Issue |
issue-handler.yml |
New Issue created | Auto-replies, detects bug Issues, and assigns them to the Copilot coding agent for automatic fixing |
deploy-pages.yml |
Push to main touching site/** |
Automatically deploys the AI-generated static site to GitHub Pages |
github-claw/
├── AGENTS.md # AI Agent identity definition and behaviour rules (entry point for every session)
├── MEMORY.md # AI long-term memory: user preferences, project background, persistent agreements
├── memory/
│ ├── tasks.md # AI task tracking (to-do / in-progress / done)
│ └── YYYY-MM-DD.md # Daily AI work log
├── .agents/
│ └── skills/ # AI Agent skill pack library (discoverable, installable, reusable)
│ ├── ui-ux-pro-max/ # UI/UX design system skill pack
│ ├── ai-image-generation/ # AI image generation skill pack
│ └── seo-audit/ # SEO audit skill pack
└── .github/
└── workflows/ # AI automation workflows (Cesium PR daily digest, Pages deploy, Issue handling)