README.md

metadata

license: apache-2.0
pretty_name: Pencil2D computer-use environment from a one-line request (REA)
tags:
  - computer-use
  - cua
  - gui-agent
  - rl-environment
  - agentenv
  - desktop
  - linux
  - rea
  - animation
  - 2d-animation
  - drawing
  - from-request

Pencil2D: a computer-use environment built from a one-line request

The real Pencil2D 0.6 animation app running on an Ubuntu 24.04 desktop, as an agentenv computer environment. An agent completes the task only through screenshots, mouse and keyboard, and a verifier grades the saved files with no model.

Claude Code built this environment from a one-line request in about 19 minutes, with only REA's build skill and MCP tools; How it was made has the request and the steps. Pencil2D opens a short bouncing-ball animation, Documents/bounce.pclx, at launch: a three-frame ball on a layer still called Bitmap Layer, a Ground layer, and Pencil2D's camera and vector layers. The agent has to rename Bitmap Layer to Ball and save. It sees only the screen and acts only with the mouse and keyboard.

Watch

• Agent runs (Claude Code with Claude Sonnet 5.5): run 1, run 2, run 3
• Witness, which proves the task can be done through the computer: witness.mp4

Each video is the computer's own screen recording at 10 frames per second.

The task

In Pencil2D, rename the "Bitmap Layer" layer of my bouncing ball animation in Documents/bounce.pclx to Ball and save it.

bundle/artifacts/rename-ball-layer-check/verify.py reads the app's files through the computer's data/get view and passes only if every check holds:

Check	Kind	File
layer-renamed	value_equals	Documents/bounce.pclx
old-name-gone	value_absent	Documents/bounce.pclx
ground-kept	value_equals	Documents/bounce.pclx
bitmap-layer-count	member_count	Documents/bounce.pclx

A .pclx file is a zip, and REA reads main.xml inside it, which lists the layers. The checks find layers by their id: layer 3 must be named "Ball", no layer may still be named "Bitmap Layer", layer 4 must keep its name, Ground, and the animation must still have two bitmap layers (member_count on type 1). Every check came from REA's draft, unedited; Claude Code discarded one more suggestion, on which layer was selected, which is view state.

Results

3 of 3 runs passed, about $0.64 per run.

Run	Outcome	Turns	Cost	Video
1	passed	10	$0.98	video
2	passed	10	$0.49	video
3	passed	10	$0.45	video

Run it

You need Docker. agentenv-framework builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (pencil2d 0.6.6).

pip install agentenv-framework==0.9.1431
hf download earakely-scale/Pencil2DFromRequestComputerEnv-REA --repo-type dataset --local-dir pencil2d-request
agent-env run pencil2d-request/bundle --task rename-ball-layer --agent claude-code --model <model>

The agent needs a model provider configured for agentenv. --task smoke only boots the computer, with no agent. bundle/README.md describes the bundle in detail.

How it was made

Claude Code built this environment from a one-line request, given only REA's build skill and REA's MCP tools:

Build me a Pencil2D environment where the agent renames a layer in a small animation and saves it.

It asked only the skill's upfront questions, wrote no app-specific scripts, and finished in about 19 minutes. The steps were REA's usual ones:

1. Capture. REA launched Pencil2D on a Linux desktop and recorded its package, windows and opened files while the edit was made. It also recorded the app's state folders before and after.
2. Export. REA turned the capture into an agentenv computer_app. It installs the app, seeds the starting files, starts the app, and waits for its main window. It also serves typed views of the app's files for verifiers.
3. Task. The task's checks were validated against the capture: they fail on the starting state and pass on the captured end state.
4. Witness. In a fresh computer, the layer was renamed through computer actions alone, and the checks were read at four stages. They held on the first try:

Stage	Checks
seed	fail, as expected: layer 3 is still named Bitmap Layer; the kept checks pass
after_edit	pass
near_miss	fail, as expected: the layer was renamed "Bal"
after_reset	fail, as expected: data/reset restored the seed

The agent runs came after the build: Claude Code with Claude Sonnet 5.5 (--model anthropic/claude-sonnet-5-5 --no-emulate, on an amd64 host) solved the task 3 times, each run recorded.

What's inside

bundle/      the agentenv bundle: envs/ (app module, seed files), tasks/, artifacts/ (verifier), evals/
media/       witness and agent-run recordings (MP4) with short GIF previews
results.json the runs above, with each check's result

Limits

• The renamed layer is found by its id, 3: a new layer named Ball in its place gets another id and fails. The drawings, the frames and the camera and vector layers' names are not checked.
• Pencil2D's settings (~/.config/Pencil) are not part of the seed, so it starts with its defaults.
• The computer installs whatever pencil2d Ubuntu 24.04's archive holds at build time (0.6.6 for these runs); it is not pinned.
• Three runs is a small sample: the solve rate shows the task is solvable, not how hard it is.

Licences

The bundle code, the task and the synthetic seed files are Apache-2.0. Pencil2D itself is not included: the computer installs it from Ubuntu's archive (pencil2d 0.6.6) at build time, under its own licence (GPL-2.0-only).