bundle/README.md

TreeLine computer app

Exported by REA's export-desktop-application from a capture of the treeline package (capture Evidence ev_cedd8aa6c80e75efe1fcd13be6c7886a6a3768a14213365580484c5b9b70d481). It was checked against agentenv-framework 0.9.1431.

• envs/treeline/ is a computer app on the ubuntu-24.04 base with the reference driver, for amd64 and arm64. It installs treeline from the distribution archive, launches sh -c cd "$HOME" 2>/dev/null; exec 'treeline' "$HOME"/'Documents/reading-list.trln' at boot and is ready when a window of class treeline.py appears.
• At launch the app opens ~/Documents/reading-list.trln, the seed's one document.
• Owned paths: ~/Documents. data/get returns their files and a typed JSON view of each; data/reset restores the seed.
• Seed: 1 file(s) made for the task, as saved at checkpoint seed. Text files and dconf keys spell the capture host's name as localhost, user@host as user@localhost and the home as /home/user, which the computer respells to its own at boot. 1 gzip file(s) were compressed again only to clear their headers, their content unchanged.
• Neither owned nor seeded, so the app starts without what it saved there: these files or folders, which changed during the capture, ~/.config/QtProject.conf, ~/.treeline-3.1. To own them, capture again naming them with state_paths on start_build_capture, or --state-path on capture-desktop-application.
• tasks/smoke.json deploys the app.

Run the task with agent-env run <this folder>, or only the deploy with --task smoke.

The seed was made for this task, not taken from a user's own data; read envs/treeline/seed/ before publishing the bundle.

Task add-history-book

tasks/add-history-book.json deploys the app, has the agent work through the computer from the prompt below, and grades the outcome with artifacts/add-history-book-check/verify.py, which reads the app's data/get views and needs no model. It grades once the app's files have held for 3 s, so a save the app makes on its own within that time of the agent stopping counts; each result's settled is false when they were still changing 15 s after the first read. evals/treeline.toml names it, so a run of the bundle runs it.

In TreeLine, add the book The Silk Roads by Peter Frankopan, published in 2015, under the History category of my reading list in Documents/reading-list.trln, and save it.

Checks, each of which must pass:

• history-gains-a-book: {"kind":"member_count","file":"Documents/reading-list.trln","pointer":"/nodes/[uid=7584d494d9f511f1a873edd4f1920549]/children","equals":3}
• one-more-book: {"kind":"member_count","file":"Documents/reading-list.trln","pointer":"/nodes","matching":"[format=BOOK]","equals":8}
• history-kept: {"kind":"value_equals","file":"Documents/reading-list.trln","pointer":"/nodes/[uid=7584d494d9f511f1a873edd4f1920549]/data/NAME","value":"History"}
• no-new-category: {"kind":"member_count","file":"Documents/reading-list.trln","pointer":"/nodes","matching":"[format=CATEGORY]","equals":3}

REA kept the task because its checks fail on the seed (history-gains-a-book, one-more-book) and all pass on the captured end state.

To measure a solve rate, run the task k times (3 by default) with the agent and model to measure, and count the runs whose summary line says passed; a run that says failed did not finish and is run again rather than counted:

agent-env run <this folder> --task add-history-book --agent claude-code --model MODEL