AgentStack
SKILL verified MIT Self-run

Gui Tool

skill-zachrouan-agent-desktop-interface-gui-tool · by ZachRouan

Interact with the desktop GUI — take screenshots, list/raise windows, click with grid targeting, type text, press key combos. Use when you need to see the screen, find windows, click on things, type into applications, or automate any GUI interaction. All commands return JSON.

No reviews yet
0 installs
20 views
0.0% view→install

Install

$ agentstack add skill-zachrouan-agent-desktop-interface-gui-tool

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

View the full security report →

Verified badge

Passed review? Show it. Paste this badge into your README — it links to the public security report.

AgentStack Verified badge Links to your public security report.
[![AgentStack Verified](https://agentstack.voostack.com/badges/verified.svg)](https://agentstack.voostack.com/security/report/skill-zachrouan-agent-desktop-interface-gui-tool)

Reliability & compatibility

Security review passed
0 installs to date
no reviews yet
18d ago

Declared compatibility

Claude CodeClaude Desktop

Compatibility is declared by the source manifest. End-to-end runtime verification is coming — see below.

Preview Execution monitoring

We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps — measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.

How agent discovery & health will work →
Are you the author of Gui Tool? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

gui-tool

Desktop GUI automation CLI. All commands return JSON to stdout (errors to stderr).

gui-tool screenshot --output /tmp/screen.png                              # full screen
gui-tool screenshot --window-id 123 --output /tmp/app.png                 # cropped to window
gui-tool screenshot --window-id 123 --grid --output /tmp/grid.png         # grid overlay
gui-tool screenshot --window-id 123 --grid --cell B2 --output /tmp/z.png  # zoom into cell

gui-tool windows list                                # JSON list of all windows
gui-tool windows raise 123                           # bring window to front

gui-tool mouse click                                 # click at current position
gui-tool mouse click --button right                  # right-click
gui-tool mouse click --cell B2.C1 --window-id 123   # click cell (recursive zoom OK)

gui-tool key type "hello" --window-id 123            # type text
gui-tool key press "ctrl+a" --window-id 123          # key combo (ctrl/shift/alt/super + key)

Grid Targeting

No pixel coordinates exist. The grid is the only way to click. Each cell has a red crosshair (+) at its center — a click on that cell lands exactly on that crosshair, nowhere else. If the crosshair isn't on your target, the click will miss. Zoom until a crosshair sits directly on the target.

Orient → Zoom → Click → Verify

Orient. Take a grid screenshot. Identify which cell contains your target and note where within that cell it sits (e.g. "the button is in D1, near the bottom-left").

Zoom. Crop into that cell to get a sub-grid. Pick the sub-cell matching your position note — don't re-read text or hunt for icons in zoomed views, they'll be blurry. Translate spatially: "bottom-left of D1" → sub-cell B5 or C5. Keep zooming (append with dots: D1.C5.F3) until a crosshair is on the target. 2–3 levels is typical.

Click. gui-tool mouse click --cell D1.C5 --window-id 123

Verify. Zoom into the area you just clicked to check the result — a full-page screenshot is too small to see subtle state changes (button color, focus ring, selection highlight). If the click missed, re-orient from scratch — the screen state may have changed.

Navigating Without Zooming Out

Zoomed views show dimmed context from adjacent parent cells with their labels visible. If your target is in a neighboring cell, reference it directly — do not zoom all the way out:

# Zoomed into G1, but target is actually in H1 (visible in dimmed context)
# WRONG: take a fresh full-grid screenshot and start over
# RIGHT: just use H1 directly
gui-tool screenshot --window-id 123 --grid --cell H1 --output /tmp/zoom.png

Only take a fresh full-grid screenshot (no --cell) when the screen has changed (after clicking, typing, or switching windows) or when you're genuinely lost.

Between-Cell Targeting

Target straddles two cells? Use + to center on the boundary:

gui-tool mouse click --cell D3+E3 --window-id 123   # horizontal
gui-tool mouse click --cell D3+D4 --window-id 123   # vertical
gui-tool mouse click --cell D3+E4 --window-id 123   # diagonal

Small Icons and Buttons

Tiny targets (like/dislike buttons, close icons, checkboxes) need 3+ zoom levels. On the first zoom, note the icon's position carefully — at deeper zooms it becomes a few blurry pixels. Trust your spatial note, not what the zoomed crop looks like.

Dot Notation

B2.C1 = sub-cell C1 within parent B2. Change the last segment to try a neighbor at the same depth (B2.D2). Append to go deeper (B2.C1.F3).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.