voice-browser-desktop-smoke-test

name	Voice Browser Desktop Smoke Test
description	Repeat Raul's quick voice-driven Prometheus smoke test: open X search for an AI-related topic, scroll it, focus Codex, then focus Claude, and only report after the tool-backed sequence completes.

---

name: Voice Browser Desktop Smoke Test description: > Runs Raul's quick end-to-end voice/control smoke test across browser and desktop surfaces. Use this when Raul says phrases like "do the test again," "voice test again," "open X search then focus Codex and Claude," "test browser and desktop focus," or asks Prometheus to verify that voice-triggered tool execution can open X, scroll, switch to Codex, and switch to Claude. This is for execution testing, not for researching AI news deeply or posting on X. version: 1.0.0 triggers: do the test again, voice test again, open X search and focus Codex, open X search then focus Claude, test browser and desktop focus, X search AI scroll Codex Claude, Prometheus voice tool test, browser desktop smoke test, test X Codex Claude, repeat the X search test

Voice Browser Desktop Smoke Test

Use this when Raul wants the quick voice-driven tool execution test that crosses browser automation and desktop focus.

Goal

Prove that Prometheus can hear Raul's voice command, execute real tools, and move through the expected surfaces without narrating fake progress:

1. Open X search for an AI-related query.
2. Scroll down a little.
3. Focus the Codex desktop app/window.
4. Focus the Claude desktop app/window.
5. Report briefly after the sequence is actually done.

Default Workflow

1. Read relevant skills first if not already loaded:
   • x-browser-automation-playbook
   • desktop-automation-playbook
2. Open X search:
   • Use browser_open("https://x.com/search?q=AI&f=live", observe:"screenshot").
   • If Raul gave a specific topic, URL-encode that query instead of AI.
3. Scroll X once:
   • Use browser_scroll({ direction:"down", multiplier:1, observe:"screenshot" }).
   • Do not collect/analyze tweets unless Raul asks.
4. Focus Codex:
   • Use desktop_find_window({ name:"Codex" }) if needed.
   • Use desktop_focus_window({ name:"Codex" }).
   • Trust the auto/fresh screenshot or capture one if focus is ambiguous.
5. Focus Claude:
   • Use desktop_find_window({ name:"Claude" }) if needed.
   • Use desktop_focus_window({ name:"Claude" }).
   • Verify visually when possible.
6. Final reply should be short and concrete, e.g.:
   • Done — opened X AI search, scrolled it, focused Codex, then focused Claude.

Rules

• This is a test flow, so skip declare_plan unless Raul explicitly asks for a plan.
• Do not say generic filler like "checking available skills" or "working in the browser."
• Do not post, like, reply, or interact socially on X.
• Do not read or alter Codex/Claude chat contents unless Raul explicitly asks.
• If Codex or Claude is not open, report the exact missing window and continue with any remaining available step.
• Prefer action first, then one brief completion message.

Non-Triggers

Do not use this skill for:

• General X research or tweet collection.
• Posting/replying/liking on X.
• Codex dev handoff prompts.
• Claude message drafting.
• Browser-only website automation unrelated to Raul's smoke test.

Quality Check

A successful run has visible/tool evidence that:

• X search opened to an AI-related query.
• The X page moved after a scroll.
• Codex was focused at least once.
• Claude was focused last.
• The final user-facing message happens only after the actual tool sequence completes.