agentleFS
Sign inStart free
naddrzznaddrzz/nalarupa/AGENTS.md

nalarupa

AGENTS.md8 starsChanged 27 days agoMITView on GitHub

AI design editor: a Rust core (core/, a cdylib loaded through bun:ffi) and a GPUIX/React UI run by Bun (app/). Windows is the release platform. CI, nightly QA and the Claude automation run on GitHub-hosted runners (Linux, plus Windows for CI); only auto-merge runs on a small self-hosted VPS runner (.github/workflows/). Work is tracked in GitHub issues. Measure a change to what the design agent is told (core/guide/*.md, its instructions in core/src/agent.rs)

  • Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.

    Of the AGENTS.md files we read, the 31,439 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.

  • In this repository2 other files it gives its agents.

The file

We split this AGENTS.md into the rules it gives an agent

2 rules. About 0 are common, 0 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Never push to dev or main.

    line 10
  2. Very specific:

    Never hold engine() across slow work: render, poll and boxes share its mutex and the UI freezes.

    line 28

≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “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
AGENTS.md
<!-- bmad:context --><!-- Verified 2026-10-18 against c7a6a33. Managed by bmad-project-context; edits inside this block are replaced on refresh. Keep anything you want preserved outside the markers. -->

nalarupa

AI design editor: a Rust core (core/, a cdylib loaded through bun:ffi) and a GPUIX/React UI run by Bun (app/). Windows is the release platform. CI, nightly QA and the Claude automation run on GitHub-hosted runners (Linux, plus Windows for CI); only auto-merge runs on a small self-hosted VPS runner (.github/workflows/). Work is tracked in GitHub issues.

Policy

  • Never push to dev or main; open a PR against dev. main is release-only and moved by the maintainer.
  • Public repo: never name, quote, or copy code, assets, kits or text from other design tools. The .pen compatibility literals already in code (pencil: URL prefix, "pencil:pending", the pencil script alias, data-pen-node-clipboard) are the only exception. Research notes stay outside the repo, never in docs/.
  • Regenerate README screenshots only with bun run shots (it uses a fresh NALARUPA_DATA, so no personal recents or chats leak).

Where things are

  • core/guide/*.md is runtime content compiled into the in-app agent (core/src/guide.rs), not instructions for you; edit it only to change what that agent is told.
  • core/vendor/resvg is resvg 0.47 with three patches and core/vendor/usvg is usvg 0.47 with one (search Nalarupa:); bump resvg, usvg and krilla-svg together and re-apply them. core/vendor/tiny-skia is tiny-skia 0.12 with one patch (Nalarupa: in pipeline/highp.rs); re-apply it when bumping.

Running and verifying

  • Build checks against your own core: set CARGO_TARGET_DIR to a separate folder and NALARUPA_CORE=<that dir>/release/nalarupa_core.dll (libnalarupa_core.so on Linux); a running app or e2e locks core/target/release/nalarupa_core.dll (link error LNK1104).
  • On Linux run GUI checks as bash scripts/wl-run.sh bun run e2e from app/; Xvfb does not work, GPUIX's Linux build has no X11 backend.
  • On Linux a11y-check, window-check, exe-check are Windows-only; name the skipped checks in the PR.
  • On Windows a real mouse moving over the background e2e window cancels simulated drags; rerun a step that "dropped early" with the cursor idle.

Conventions that differ from defaults

  • Never hold engine() across slow work: render, poll and boxes share its mutex and the UI freezes; do heavy work after the guard drops (see ToolOut::finish, core/src/engine.rs).
  • Set rquickjs runtime handlers (set_interrupt_handler) outside Context::with; inside it they deadlock (feature parallel).
  • GPUI does not bubble clicks: give a small clickable child a painted fill ('#00000001'), not transparent, so it gets its own hitbox.
  • Draw focus rings with FocusRing (app/src/ui.tsx), never boxShadow, which GPUI paints as a filled box.
  • Keys reach only the focused element's onKeyDown; window-level keys go in the render(<App/>, { onKeyDown }) options in app/app.tsx.

Known pitfalls

  • GPUIX automation: chords use hyphens (ctrl-a; ctrl+a silently does nothing); fill() hangs on text containing double quotes; <input> values and style cannot be read, so assert on visible text or bounds; synthetic clicks carry no clickCount (double-click handlers use a 450 ms fallback).
  • GPUIX fill() can miss an <input> inside <virtual-list> (an orphan focus handle); click the field, then send keystrokes without an elementId (see editor-e2e). Under the slow Linux renderer, poll with until() instead of fixed sleeps.
  • Give every <textarea> an aria-label; its placeholder is not its screen-reader name and a11y-check fails.
  • Match each file's line endings when patching by string; the repo mixes CRLF and LF.
<!-- /bmad:context -->

Changing the design agent

Measure a change to what the design agent is told (core/guide/*.md, its instructions in core/src/agent.rs) with an eval run before and after: cd app && bun run design-eval run, then bun run design-eval compare <before> <after> (README Checks).

Agent skills

Issue tracker

Issues live in this repo's GitHub Issues (gh CLI). See docs/agents/issue-tracker.md.

Triage labels

The five default triage labels (needs-triage, needs-info, ready-for-agent, ready-for-human, wontfix). See docs/agents/triage-labels.md.

Domain docs

Single-context: one GLOSSARY.md and docs/adr/ at the repo root. See docs/agents/domain.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.

Check your setup, free