earakely-scale/ScribusFromRequestComputerEnv-REA

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

The real Scribus 1.6 desktop publishing 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 24 minutes, with only REA's build skill and MCP tools; How it was made has the request and the steps. Scribus opens a one-page event flyer, Documents/jazz-night-flyer.sla, at launch. The agent has to change its headline from Riverside Jazz Night to Harbor Lights Jazz Festival and save, leaving the details line below it alone. 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

Change the headline of the event flyer in Documents/jazz-night-flyer.sla to Harbor Lights Jazz Festival and save it.

bundle/artifacts/change-flyer-headline-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
headline	value_equals	Documents/jazz-night-flyer.sla
details-untouched	value_equals	Documents/jazz-night-flyer.sla

A Scribus .sla file is XML, and a text frame keeps its text in the CH attribute of an ITEXT element. The checks find each frame by the name it was given (ANNAME Headline and Details), never by position: the headline must read "Harbor Lights Jazz Festival" exactly, and the details line must still read "Saturday, June 14 at 7:00 PM - Maple Street Community Hall". REA's draft picked the headline frame by its x coordinate; Claude Code switched it to the frame's name, so moving the frame does not break the check.

Results

3 of 3 runs passed, about $0.62 per run.

Run	Outcome	Turns	Cost	Video
1	passed	9	$0.94	video
2	passed	8	$0.67	video
3	passed	12	$0.25	video

Run it

You need Docker. agentenv-framework builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (scribus 1.6.1).

pip install agentenv-framework==0.9.1431
hf download earakely-scale/ScribusFromRequestComputerEnv-REA --repo-type dataset --local-dir scribus-request
agent-env run scribus-request/bundle --task change-flyer-headline --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 Scribus environment where the agent changes the headline on a one-page event flyer and saves it.

It asked only the skill's upfront questions, wrote no app-specific scripts, and finished in about 24 minutes. The steps were REA's usual ones:

1. Capture. REA launched Scribus 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 edit was made through computer actions alone, and the checks were read at four stages. They held on the first try:

Stage	Checks
seed	fail, as expected
after_edit	pass
near_miss	fail, as expected: the headline was retyped as "Harbor Lights Jazz Festval"
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 check expects the headline to stay one ITEXT run. A correct headline whose characters were styled differently, which splits it into several runs, fails: REA's whole-text checks join text nodes, not text kept in attributes.
• The seed also holds Scribus's font cache, .config/scribus/checkfonts150.xml, which the capture picked up; it lists only the computer's system fonts.
• The computer installs whatever scribus Ubuntu 24.04's archive holds at build time (1.6.1 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. Scribus itself is not included: the computer installs it from Ubuntu's archive (scribus 1.6.1) at build time, under its own licence (GPL-2.0-or-later).