agentleFS
oliver-kriskaoliver-kriska/prmarmot/.claude/skills/verify-ui/SKILL.md

verify-ui

Skill10 starsChanged 40 days agoMITView on GitHub

Screenshot-verify what the running prboard GPUI window actually renders. Use this EVERY time visual confirmation is needed — after any UI/theme/layout change, before claiming a visual fix works, when checking light vs dark mode, or when a screenshot looks stale or the app looks "hung". Never verify prboard visuals with a bare screencapture or a full-screen shot; this skill exists because the naive approaches silently capture wrong or stale pixels.

9 other projects use a version of this.

Adopted by

Most recently taken up by

  • Other versions9 projects use a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 9

    1. ThisIsSadeghi/KMPilot★ 56same name
    2. zillydev/claude-code-manager★ 4same name
    3. jonmumm/skills★ 3same name
    4. fimoklei/maestro★ 1same name
    5. rpraharaj/ai-engineer★ 1same name
    6. ThisIsSadeghi/Loupe★ 1same name
    7. crsmithdev/aleph★ 0same name
    8. rpraharaj/ai-sdlc★ 0same name
    9. whyashthakker/estack★ 0same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository2 other files it gives its agents.

The file

Files
SKILL.md

Verify prboard UI by screenshot

Why the naive approach fails (all three learned the hard way)

  1. GPUI does not repaint occluded windows. A window behind another window (or on another Space) keeps its last frame — often the first frame, which shows "Loading board…" and looks exactly like a hung fetch. The app is fine; the pixels are stale. The window must be frontmost before capture.
  2. Multiple prboard instances share one app name. A dev binary, the installed .app, and a measurement instance can all run at once; capturing "the prboard window" by name grabs an arbitrary one. Always select the window by PID.
  3. Keystroke automation is dangerous. Sent keys land in whichever window has focus at that instant; a focus race once delivered q to the memory-gate instance and killed an overnight measurement. While a gate run is live (check for a measure.sh process), send no keystrokes at all. Outside gate runs, re-activate the target window immediately before each keystroke, never once for a batch.

Procedure

  1. Find the PID of the instance you mean:
    pgrep -lf prboard    # distinguish target/debug, target/release, and prboard.app paths
  2. Capture (activates the window by PID, resolves its CGWindowID, screenshots just that window):
    .claude/skills/verify-ui/scripts/capture-window.sh <pid> "$SCRATCHPAD/ui.png"
  3. Read the PNG with the Read tool and actually look at it. Check against DESIGN.md (palette, dots-not-emoji, calm rule) — not just "did it render".
  4. For theme comparisons, capture once per theme; for before/after, keep both files and name them (before-*.png / after-*.png).

On demo data, make capture (scripts/capture.sh) does all of this for the board, review queue, All open and a Details panel, in both themes. To press keys in an instance you launched, post them to its PID rather than typing into the focused window: swiftc -O scripts/capture/post-event.swift -o "$SCRATCHPAD/post-event", then "$SCRATCHPAD/post-event" <pid> key:125 key:49 (Down, Space). Posted clicks do not make GPUI clicks, so a scene behind a button needs a key.

Launching a throwaway instance for verification is fine (./target/debug/prboard --repo owner/name) — quit it with a mouse click on close or kill <pid>, never a synthesized q, and leave any measured instance untouched.

Needs macOS Screen Recording permission for the terminal (screencapture silently produces empty/desktop images without it) and swift on PATH (ships with Xcode).

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.