Ray Skill: PHP-First Debugging (Curl Edition)

This skill integrates Spatie's Ray with the AI agent using curl to interact with the Ray MCP. It prioritizes a fast, tool-independent debugging loop.

Primary Workflow

1. Inject & Prepare

Insert the ray() call into the relevant PHP file.

• Tools: Use replace or write_file.
• Snippet: ray($var)->label('Debug Point'); or ray()->trace();.
• Cache Warning: In environments like Magento, you MUST flush relevant caches (e.g., bin/magento cache:flush layout full_page) after injection.

2. Execute

Trigger the code execution.

• Web: Refresh the browser or use curl -k -I <URL>.
• CLI: Run the relevant command or script.

3. Discover (One-time per session)

Find the active Ray window to target the correct project.

curl -s -X POST http://localhost:2411/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc": "2.0", "method": "tools/call", "params": {"name": "get_ray_windows", "arguments": {}}, "id": 1}'

Note the projectName and hostname from the result.

4. Capture Logs

Retrieve the logs using the discovered project name.

curl -s -X POST http://localhost:2411/mcp \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc": "2.0", "method": "tools/call", "params": {"name": "get_ray_window_logs", "arguments": {"projectName": "default", "limit": 5}}, "id": 1}'

5. Cleanup

Remove injected ray() calls immediately after verification.

MCP Tool Reference (via Curl)

• get_ray_windows: Lists active projects and hostnames.
• get_ray_window_logs: Retrieves logs (requires projectName).
• send_ray_clear_all: Clears the window.
• send_custom_html_output: Sends rich HTML to Ray.

Cache-Heavy Environments (e.g. Magento)

Always assume the template is cached.

1. Inject ray().
2. bin/magento cache:flush layout full_page.
3. Trigger page load.
4. Capture logs.
5. Revert changes + bin/magento cache:flush.