docs(claude): note Claude-in-Chrome browser automation for real-UI validation

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

CLAUDE.md

@@ -64,6 +64,14 @@ All shippable work commits **directly to `main`**; `main` is always equal to wha
 
 Live validation of a user-facing feature must exercise the real UI/user flow end-to-end (e.g. connect → enroll → deploy), not shortcut via direct API/agent/CLI calls — a shortcut can give false confidence and leave the system inconsistent (see the F9 storage episode). Low-level mechanism tests where the direct call IS the mechanism are exempt.
 
+**Browser automation is available** for that real-UI validation: the Claude Code↔Chrome bridge
+(`claude-in-chrome`) is installed, so CC can drive the dashboard / FileBrowser / app admin UIs at
+`*.demo-felhom.eu` (click Telepítés, upload via FileBrowser, register an immich library, etc.) and
+screenshot evidence. **Caveat:** the browser tools attach only to a session STARTED AFTER the bridge
+connected — if they're missing, open a fresh task (don't infer the feature is broken). `/mcp` should
+list `claude-in-chrome · connected`; otherwise `/chrome` → reconnect. (CC env note: the long-installed
+claude.ai host is a DIFFERENT bridge — Claude Code needs `com.anthropic.claude_code_browser_extension`.)
+
 ## Environment
 
 | Machine | OS | IP | Purpose |