Initial commit
This commit is contained in:
Zhongwei Li
54
commands/browser.md
Normal file
54
commands/browser.md
Normal file
@@ -0,0 +1,54 @@
|
||||
---
|
||||
description: Start/restart the Browser MCP server for automated web operations
|
||||
---
|
||||
|
||||
# Browser MCP Server Management
|
||||
|
||||
Start or restart the Browser MCP server for automated web operations.
|
||||
|
||||
## Task
|
||||
Restart the Browser MCP server on port 9222. If already running, stop it first to ensure new environment settings are loaded.
|
||||
|
||||
## Steps
|
||||
|
||||
1. Stop any existing Browser MCP server processes:
|
||||
```bash
|
||||
pkill -f "ts-node browser-server.ts" || true
|
||||
```
|
||||
|
||||
2. Wait for clean shutdown:
|
||||
```bash
|
||||
sleep 2
|
||||
```
|
||||
|
||||
3. Start the server in the background:
|
||||
```bash
|
||||
cd orchestra/mcp-servers && npm run browser &
|
||||
```
|
||||
|
||||
4. Wait for startup and verify:
|
||||
```bash
|
||||
sleep 3 && curl -s http://localhost:9222/health | jq
|
||||
```
|
||||
|
||||
5. Show usage instructions:
|
||||
```
|
||||
✅ Browser MCP server is running on http://localhost:9222
|
||||
👁️ Browser mode: GUI visible (set BROWSER_HEADLESS=true for headless mode)
|
||||
|
||||
Quick test:
|
||||
./orchestra/mcp-servers/browser-helper.sh init
|
||||
./orchestra/mcp-servers/browser-helper.sh navigate https://example.com
|
||||
./orchestra/mcp-servers/browser-helper.sh screenshot example.png true
|
||||
./orchestra/mcp-servers/browser-helper.sh close
|
||||
|
||||
See: orchestra/agents/skills/web-browse.md for full documentation
|
||||
```
|
||||
|
||||
## Notes
|
||||
- The server always restarts to pick up new .env settings
|
||||
- All browser operations are logged to artifacts/browser/{sessionId}/
|
||||
- Rate limits: 10 navigations, 50 clicks, 30 inputs per session
|
||||
- Default port: 9222 (Chrome DevTools Protocol standard port)
|
||||
- Default mode: GUI visible (browser window opens automatically)
|
||||
- To stop: `pkill -f "ts-node browser-server.ts"`
|
||||
77
commands/orchestra-setup.md
Normal file
77
commands/orchestra-setup.md
Normal file
@@ -0,0 +1,77 @@
|
||||
---
|
||||
description: Run the Orchestra Plugin setup script to install all MCP servers and dependencies
|
||||
---
|
||||
|
||||
# Orchestra Plugin Setup
|
||||
|
||||
Run the complete setup script to install all MCP servers, dependencies, and configure the Orchestra Plugin environment.
|
||||
|
||||
## Task
|
||||
|
||||
Execute the Orchestra setup script to install:
|
||||
- Node.js dependencies (Express, Playwright, TypeScript)
|
||||
- Playwright Chromium browser
|
||||
- Python virtual environment and packages (elevenlabs, requests)
|
||||
- Configure hooks and permissions (.claude/settings.json with auto-approve)
|
||||
- Install safety guard hook (blocks dangerous operations only)
|
||||
- Create necessary directories (artifacts, etc.)
|
||||
|
||||
## Steps
|
||||
|
||||
1. Check if setup.sh exists and is executable:
|
||||
```bash
|
||||
if [ ! -f "./setup.sh" ]; then
|
||||
echo "❌ setup.sh not found. Please ensure you're in the Orchestra repository root."
|
||||
exit 1
|
||||
fi
|
||||
```
|
||||
|
||||
2. Make setup.sh executable if needed:
|
||||
```bash
|
||||
chmod +x ./setup.sh
|
||||
```
|
||||
|
||||
3. Run the setup script:
|
||||
```bash
|
||||
./setup.sh
|
||||
```
|
||||
|
||||
4. After completion, show next steps:
|
||||
```
|
||||
✅ Orchestra Plugin setup completed!
|
||||
|
||||
Next steps:
|
||||
1. Edit .env file to add your API keys (if not done already)
|
||||
2. Restart Claude Code to activate all features
|
||||
3. Start coding - Orchestra will automatically enhance your workflow!
|
||||
|
||||
Available features:
|
||||
- /browser - Start Browser MCP server
|
||||
- /screenshot - Capture web screenshots
|
||||
- 11 specialized AI agents (Riley, Skye, Finn, Eden, Kai, Leo, Iris, Nova, Mina, Theo, Blake) coordinated by main Claude Code
|
||||
- Automated quality gates (before_task, before_pr, before_merge, before_deploy, after_deploy)
|
||||
- Multi-agent orchestration with parallel execution
|
||||
```
|
||||
|
||||
## Notes
|
||||
|
||||
- This command should be run once after cloning the Orchestra repository
|
||||
- The setup script checks for prerequisites (Node.js 18+, Python 3.8+)
|
||||
- All hooks automatically skip if tools aren't installed (no errors)
|
||||
- Environment variables can be configured later in the .env file
|
||||
- Only GITHUB_TOKEN is required; all other tokens are optional
|
||||
- The script creates symlinks for hooks and slash commands
|
||||
- Browser server can be started separately with `/browser` command
|
||||
|
||||
## Auto-Approval & Safety
|
||||
|
||||
The setup creates `.claude/settings.json` with:
|
||||
- `"permissions.allow": ["*"]` - Auto-approve all tools (no confirmation dialogs)
|
||||
- Safety guard hook (`hooks/user-prompt-submit.sh`) blocks dangerous operations:
|
||||
- `rm -rf /`, system file modifications
|
||||
- `sudo shutdown/reboot`
|
||||
- `git push --force`
|
||||
- Database drops (`DROP DATABASE/TABLE`)
|
||||
- And more...
|
||||
|
||||
This enables autonomous operation while preventing destructive actions.
|
||||
56
commands/screenshot.md
Normal file
56
commands/screenshot.md
Normal file
@@ -0,0 +1,56 @@
|
||||
---
|
||||
description: Capture screenshots from browser automation
|
||||
---
|
||||
|
||||
# Take Screenshot
|
||||
|
||||
Capture a screenshot of a web page using the Browser MCP server.
|
||||
|
||||
## Task
|
||||
Navigate to a specified URL and take a full-page screenshot, saving it to the artifacts directory.
|
||||
|
||||
## Steps
|
||||
|
||||
1. Ask the user for the URL if not provided in the command arguments
|
||||
|
||||
2. Check if Browser MCP server is running:
|
||||
```bash
|
||||
curl -s http://localhost:3030/health
|
||||
```
|
||||
If not running, tell the user to run `/browser` first
|
||||
|
||||
3. Generate a unique filename based on the URL (e.g., `example-com-{timestamp}.png`)
|
||||
|
||||
4. Execute the screenshot workflow:
|
||||
```bash
|
||||
cd orchestra/mcp-servers
|
||||
|
||||
# Initialize browser
|
||||
./browser-helper.sh init
|
||||
|
||||
# Navigate to URL
|
||||
./browser-helper.sh navigate {URL}
|
||||
|
||||
# Wait for page load
|
||||
sleep 2
|
||||
|
||||
# Take full-page screenshot
|
||||
./browser-helper.sh screenshot {filename}.png true
|
||||
|
||||
# Close browser
|
||||
./browser-helper.sh close
|
||||
```
|
||||
|
||||
5. Show the screenshot path to the user and offer to open it
|
||||
|
||||
## Example Usage
|
||||
```
|
||||
User: /screenshot https://example.com
|
||||
Assistant: Taking screenshot of https://example.com...
|
||||
✅ Screenshot saved to: artifacts/browser/{sessionId}/example-com-1234567890.png
|
||||
```
|
||||
|
||||
## Notes
|
||||
- Requires Browser MCP server to be running (use `/browser` to start)
|
||||
- Screenshots are saved to artifacts/browser/{sessionId}/
|
||||
- Full-page screenshots capture the entire page, not just the viewport
|
||||
Reference in New Issue
Block a user