agentleFS
LeostrukaLeostruka/devin-bundle/skills/computer-use/SKILL.md

computer-use

Skill0 starsChanged 3 months agoMITView on GitHub

Use when you need to see or control the local screen (screenshots, screen recording with video + contact sheet for catching transient bugs/animations, mouse clicks/movement, typing text and hotkeys) or to read/control terminals (Windows Terminal, conhost, mintty, spawned PTYs). GUI automation for desktop apps and browsers, replicating Devin Cloud Computer Use in the CLI.

68 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use competitor-news-monitor, claude-buddy and 10 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • aivrarSkill

    competitor-news-monitor

    Watch named companies for material news; cited digests.

    Also in 6 projects that use this

  • autonomous-aiSkill

    claude-buddy

    Coordinate with the Claude Buddy companion (Claude Desktop + Claude Code on the user's Mac) — voice-approve tool permission prompts and stay aware of Claude activity.

    Also in 3 projects that use this

  • autonomous-aiSkill

    face-enroll

    Manage the device's face recognition roster — enroll new faces (3 paths: user-supplied photo, agent-captured snapshot on user request, or HAL's familiar-stranger prompt) and maintain the enrolled set (status / remove / reset). All enrolled persons are friends; strangers stay unnamed until promoted via one of the enroll flows.

    Also in 3 projects that use this

  • +9 more from autonomous-ai

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 2 are distinctive, and 1 is very specific to this project.

  • 0

    Common

    Most of the 63 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Always invoke via the extension's isolated venv Python:

    line 28
  2. Distinctive:

    Use for transient bugs, animations, flicker, and anything a single screenshot misses.

    line 62
  3. Very specific:

    If the venv is missing, run the bundle installer (install.ps1/install.sh) or create it per USAGE.md.

    line 74

“Common” is measured: kept by at least 6 in 10 of the 63 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Computer Use

Local GUI automation. The scripts live in the computer-use extension, not in this skill — this file is only the router pointer.

Where it is

  • Bundle source: extensions/computer-use/
  • Installed: %APPDATA%\devin\extensions\computer-use\ (Windows) or ~/.config/devin/extensions/computer-use/ (POSIX)
  • Full docs: USAGE.md inside that directory — read it first.

Quick start

At session start, ask the user which action profile to use (ask_user_question: fast / smooth / human), then persist it:

PY profile.py --set human              # writes session file in OS temp dir

Always invoke via the extension's isolated venv Python:

  • Windows: %APPDATA%\devin\extensions\computer-use\.venv\Scripts\python.exe
  • POSIX: ~/.config/devin/extensions/computer-use/.venv/bin/python
PY screenshot.py --hints               # Vimium-style badges on real UI elements
PY screenshot.py --grid                # fallback when UIA has no elements
PY mouse.py click --hint as            # click element center by hint id
PY mouse.py click X Y                  # or click physical pixel directly
PY type_text.py "text" | --keys ctrl+c # type / hotkey
PY record.py --seconds 5               # screen video + contact-sheet PNG
PY terminal.py bind --hwnd <n>          # read/control an existing terminal
PY terminal.py spawn --shell cmd        # or own a PTY session (daemon)
PY terminal.py link <src> <dst>         # pipe session A output into B input
PY terminal.py links | unlink <id>      # inspect / stop a link
PY browser.py bind --endpoint <url> --pid <p>  # browser via CDP/BiDi
# Android: envs/<id>.json {"provider":"adb","serial":"..."} then:
PY screenshot.py --env <id> --hints     # uiautomator tree -> hint badges
PY mouse.py click --hint as --env <id>  # tap element center via adb
PY terminal.py exec "ls /sdcard" --env <id>  # adb shell

Terminal/browser details (read paths, send/exec/recv, events daemon) are in USAGE.md — same directory.

Workflow: screenshot.py --hints → pick the target by name/type from the JSON (or the badge letters in the PNG) → mouse.py click --hint <id> → re-screenshot to verify. For visual checks (images, canvas, rendered content) use plain screenshot.py; hints only cover interactive elements.

Recording: record.py writes a video file (mp4/webm via ffmpeg, animated webp/gif via Pillow fallback) plus a contact-sheet PNG. Read the sheet with the read tool to see what happened; video can't be viewed directly. Use for transient bugs, animations, flicker, and anything a single screenshot misses.

Screenshots are disposable: keep the default temp path, and delete the shots you created when the task ends. Only write --out elsewhere if the user asks to keep one.

Isolated environments: any CLI takes --env <env_id> to act inside an approved QEMU/container guest instead of the host — lifecycle via env.py (doctor/create/start/stop/restart/reset; TTY consent required). Details and limits: USAGE.md → "Isolated environments".

If the venv is missing, run the bundle installer (install.ps1/install.sh) or create it per USAGE.md.

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.