earakely-scale/CherryTreeFromRequestComputerEnv-REA

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

The real CherryTree 1.1 hierarchical note-taking 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 27 minutes, with only REA's build skill and MCP tools; How it was made has the request and the steps. CherryTree starts with no notebook open. Its notebook, Documents/home-notebook.ctd, holds three notes, Weekend chores, Groceries and Gift ideas, and Weekend chores lists four chores. The agent has to open the notebook, add a line saying Clean the gutters to Weekend chores, 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 CherryTree, add a line saying Clean the gutters to the Weekend chores note in my notebook Documents/home-notebook.ctd and save it.

bundle/artifacts/add-chore-line-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
new-line	text_contains	Documents/home-notebook.ctd
kept-water	text_contains	Documents/home-notebook.ctd
kept-sweep	text_contains	Documents/home-notebook.ctd
kept-bulb	text_contains	Documents/home-notebook.ctd
kept-library	text_contains	Documents/home-notebook.ctd
three-notes	member_count	Documents/home-notebook.ctd

A .ctd file is CherryTree's XML notebook, with one node element per note. The checks find the Weekend chores note by its name and read its whole text: it must contain "Clean the gutters" and still contain the four existing chores, and the notebook must still hold three notes (member_count). REA's draft suggested the new-line check and the four kept-chore guards, all kept exactly as written; Claude Code added the count.

Results

3 of 3 runs passed, about $0.77 per run.

Run	Outcome	Turns	Cost	Video
1	passed	14	$0.84	video
2	passed	13	$0.57	video
3	passed	15	$0.90	video

Run it

You need Docker. agentenv-framework builds the computer: Ubuntu 24.04 with the app installed from Ubuntu's archive (cherrytree 1.1.2, with librsvg2-common).

pip install agentenv-framework==0.9.1431
hf download earakely-scale/CherryTreeFromRequestComputerEnv-REA --repo-type dataset --local-dir cherrytree-request
agent-env run cherrytree-request/bundle --task add-chore-line --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 CherryTree environment where the agent adds a line to a note in a small notebook and saves it.

It asked only the skill's upfront questions, wrote no app-specific scripts, and finished in about 27 minutes. The steps were REA's usual ones:

1. Capture. REA launched CherryTree 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 line was added through computer actions alone, after opening the notebook, and the checks were read at four stages. They held on the first try:

Stage	Checks
seed	fail, as expected: Weekend chores has no gutters line; the kept checks pass
after_edit	pass
near_miss	fail, as expected: the line was changed to "Clean the guters"
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

• CherryTree opens no notebook at launch. An REA bug gave the seed's ~/.cache/dconf/user the document role, so the export saw two documents and opened neither. The agent has to open Documents/home-notebook.ctd itself; the solvers did it with Ctrl+O.
• The checks look for the new text anywhere in the Weekend chores note, so it need not be a line of its own; the Groceries and Gift ideas notes' contents are not checked.
• The seed also carries Documents/.home-notebook.ctd~, a backup CherryTree wrote of an earlier version of the notebook; the checks read only home-notebook.ctd.
• This published copy leaves out one seed file, the 2-byte ~/.cache/dconf/user (two zero bytes), because Hugging Face refuses files under a .cache/ folder. bundle/README.md still counts it. ~/.cache/dconf stays owned, so data/reset clears whatever the app writes there; nothing the task reads depends on it.
• The bundle installs librsvg2-common, the SVG image loader CherryTree's icons need, beside the app as an extra package.
• The computer installs whatever cherrytree Ubuntu 24.04's archive holds at build time (1.1.2 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. CherryTree itself is not included: the computer installs it from Ubuntu's archive (cherrytree 1.1.2, with librsvg2-common) at build time, under its own licence (GPL-3.0-or-later).