Claude 5c0a5fa899 docs(agents): run the shared UI on the JVM to verify UI changes ...

Agents had no documented way to look at a UI change: compiling and unit
tests miss a screen that crashes on open, a missing string or a
collapsed layout. The desktop app renders the same commonsUI screens,
shell and feeds as Android, so running it is the cheap real check.

- tools/desktop-run/run-headless.sh: Xvfb + :desktopApp:run on a scratch
  profile, waits for the first paint, screenshots; logs/stop commands.
  Uses the software renderer, drops saved window bounds so the window
  fits the screen, and matches the app JVM without killing the caller.
- tools/desktop-run/Drive.java: pure-JDK java.awt.Robot driver
  (screenshot, click, type via clipboard, key chords, scroll, waitpaint),
  so no xdotool or window manager is needed.
- AGENTS.md: "Run the app before calling a UI change done".
- desktop-expert: headless run section, description trigger, and the
  test task is :desktopApp:test (plain JVM module), not jvmTest.
- compose-expert: pointer to the same loop.

Verified in a cloud container: login screen, npub login, feed, phone
tier at 480x900 with bottom bar, drawer + Escape, restart in ~18 s.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017nKjeFL13yDUDtFxEHSUrr

2026-10-12 14:13:50 +00:00

..

references

chore(skills): consolidate 33 skills into 14, verified against the tree

2026-10-11 12:21:01 -04:00

SKILL.md

docs(agents): run the shared UI on the JVM to verify UI changes

2026-10-12 14:13:50 +00:00