🦞 Lobstore Skills
A VS Code extension that allows you to browse, preview, and download AI agent skills from the ClawHub skills repository.
Similar to Awesome Copilot, but specifically designed for the openclaw/skills ecosystem with instant loading and auto-updates.
Features
- ⚡ Instant Loading: Skills data bundled with the extension for zero-latency browsing
- 🔍 Browse: Explore skills in a convenient tree view grouped by category
- 📖 Preview: View SKILL.md content before downloading
- ⬇️ Download: Save skills to
~/.copilot/skills/ for GitHub Copilot
- 🔃 Auto-Update: GitHub Action refreshes skills daily (no user action needed)
- 💾 Offline-Ready: Browse skills even without internet
How It Works
Data Source: The extension loads skill metadata from a pre-built JSON file that is updated daily by a GitHub Action. This avoids GitHub API rate limits and provides instant loading.
Updates: The skill catalog is automatically refreshed every day at 6am UTC via GitHub Actions workflow. Extension users get the latest skills when they update the extension.
How to Use
- Open the Extension: Click the 🦞 Lobstore icon in the Activity Bar
- Browse Skills: Expand categories to see available skills (loads instantly!)
- Preview Content: Click the preview icon (👁️) on any skill to see its SKILL.md
- Download Skills: Click the download icon (⬇️) to install to your system
- Manage Installed: View and uninstall skills from the "Installed" tab
Folder Structure
Downloaded skills are installed to your workspace by default:
.github/skills/ ← Workspace skills (recommended)
OR
~/.copilot/skills/ ← Global skills
└── skill-name/
├── SKILL.md ← Main skill file
├── scripts/ ← Optional helper scripts
├── references/ ← Optional reference docs
└── examples/ ← Optional examples
GitHub Copilot automatically detects skills in .github/skills/ within your workspace.
Compatibility
Skills from ClawHub work with:
| Tool |
Location |
| GitHub Copilot |
~/.copilot/skills/ |
| Claude Code |
~/.claude/skills/ |
| Codex |
~/.codex/skills/ |
| OpenClaw |
~/.openclaw/skills/ |
All these tools use the same AgentSkills open standard.
Requirements
- VS Code 1.85.0 or higher
- GitHub Copilot (optional, but recommended for using the skills)
Extension Settings
| Setting |
Default |
Description |
lobstore.skillsDirectory |
<workspace>/.github/skills |
Where to install skills (leave empty for workspace default) |
lobstore.registryUrl |
https://api.github.com/repos/openclaw/skills |
Skills repository URL |
lobstore.dataUrl |
https://raw.githubusercontent.com/garrettsiegel/lobstore/main/data/skills.json |
URL for auto-updating skills data |
Commands
| Command |
Description |
Lobstore: Refresh |
Reload skills list from bundled data |
Lobstore: Search |
Search for skills |
Lobstore: Preview |
Preview a skill's SKILL.md |
Lobstore: Download |
Install a skill |
Lobstore: Uninstall |
Remove an installed skill |
Lobstore: Security Scan |
Scan a skill for security issues before/after install |
Lobstore: Open in Browser |
View skill on GitHub |
Lobstore: Reveal in Finder |
Show installed skill in file manager |
🛡️ Security Scanning
Lobstore includes built-in security scanning that automatically checks skills before installation. This helps protect you from:
Detected Threats
| Category |
Examples |
| Destructive Commands |
rm -rf /, system wipes, fork bombs |
| Remote Code Execution |
curl ... \| bash, download-and-execute patterns |
| Credential Exfiltration |
Sending API keys, tokens, or passwords to external servers |
| Sensitive File Access |
Reading ~/.ssh, ~/.aws, or other credential files |
| Reverse Shells |
Netcat, socat, or other backdoor patterns |
| Prompt Injection |
Instructions that try to override AI safety guidelines |
| Permission Escalation |
Suspicious chmod/chown operations |
| Persistence Mechanisms |
Cron jobs, systemd services, launchd agents |
How It Works
- Automatic Scan: Every skill is scanned before installation
- Deep Scan: All extracted text/code files are scanned (not just
SKILL.md)
- Archive Guards: Blocks path traversal, executable binaries, oversized files, and archive bombs
- Blocking: Critical/high-severity issues block installation in strict mode (default)
- Warnings: Medium/low issues show a warning but allow installation
- Manual Scan: Right-click any skill → "Security Scan" to preview the report
- Bypass Option: Trusted skills can force-install with explicit confirmation
Security Score
Each skill receives a score from 0-100:
- 🟢 90-100: Safe - No significant issues
- 🟡 70-89: Caution - Minor concerns
- 🟠 50-69: Warning - Review recommended
- 🔴 0-49: Blocked - Installation prevented
Note: Security scanning is not foolproof. Always review skills from unknown authors.
Security Settings
| Setting |
Default |
Description |
lobstore.security.strictMode |
true |
Blocks install on high-risk findings |
lobstore.security.maxFilesExtracted |
500 |
File count limit for extracted archives |
lobstore.security.maxArchiveSizeMb |
25 |
Max total uncompressed archive size |
lobstore.security.maxSingleFileSizeMb |
2 |
Max allowed size of a single extracted file |
Architecture
Data Pipeline
GitHub Repo (openclaw/skills)
↓
GitHub Actions (daily cron @ 6am UTC)
↓
scripts/fetch-skills.js (fetch all skills)
↓
data/skills.json (bundled with extension)
↓
VS Code Extension (instant load)
No Rate Limits
By pre-fetching and bundling skills data, we avoid:
- ❌ GitHub API rate limits (60 req/hour unauthenticated)
- ❌ Slow initial load times
- ❌ Network dependency for browsing
Updating Skills Data
The extension data is updated automatically by:
- GitHub Action runs daily (
.github/workflows/sync-skills.yml)
- Fetches all skills from openclaw/skills repo
- Generates
data/skills.json with full catalog
- Commits back to repo
- Next extension build includes updated data
What are Skills?
Skills are instruction sets that teach AI assistants how to perform specialized tasks. Each skill contains:
- SKILL.md: Instructions the AI follows
- Scripts: Optional automation helpers
- References: Documentation the AI can reference
- Examples: Sample code or outputs
When you ask Copilot something relevant to a skill, it automatically loads those instructions.
Example: Install the frontend-design skill, then ask Copilot "build me a dashboard" - it will follow the skill's design guidelines.
Development
Setup
npm install
npm run compile
npm run copy-data
Testing
Press F5 to launch Extension Development Host
Manual Data Refresh
To manually update skills data (for development):
node scripts/fetch-skills.js
npm run compile && npm run copy-data
Project Structure
lobstore/
├── .github/
│ └── workflows/
│ └── sync-skills.yml ← Daily auto-update workflow
├── src/
│ ├── extension.ts ← Main entry point
│ ├── api/clawhub.ts ← Loads skills from JSON
│ ├── services/installer.ts ← Downloads and installs skills
│ ├── views/ ← Tree view providers
│ └── utils/ ← Path utilities
├── scripts/
│ └── fetch-skills.js ← Fetches from GitHub API
├── data/
│ └── skills.json ← Pre-built skills catalog (144 skills)
└── dist/ ← Compiled output
Credits
Note: This is an unofficial extension. It is not affiliated with GitHub, Microsoft, or Anthropic.
License
MIT
This extension is not affiliated with ClawHub, GitHub, or Anthropic.