agentleFS
TresnandaTresnanda/kybern/CLAUDE.md

kybern

CLAUDE.md83 starsChanged 58 days agoMITView on GitHub

- Work in the worktree you were started in. Never cd into the main checkout. - The desktop app is Kybern GPUIX in apps/gpuix (macOS, Windows, Linux). The Tauri app was removed (#113); the old GPUI client on the gpui branch is archived and must not be ported back. - Keep the existing look: reuse apps/gpuix/src/kit before writing new styles, change only what was asked, and screenshot the app window after UI changes. - Native GPUIX changes are commits on…

The file

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

8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    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:

    Keep the existing look: reuse apps/gpuix/src/kit before writing new styles, change only what was asked, and screenshot the app window after …

    line 9
  2. Very specific:

    If you were delegated: never push the "kybern delegation base" commit, wait for CI with blocking commands, and finish with the full report …

    line 14
  3. Distinctive:

    Every error a feature can show follows "Errors people can act on" in AGENTS.md: a plain headline, one sentence on what to do, a fix button …

    line 20
  4. Distinctive:

    Follow the Disk usage and cleanup rules in AGENTS.md for every build and temporary worktree.

    line 28
  5. Distinctive:

    Check for active processes/open files first.

    line 37
  6. Distinctive:

    Treat fallback text and hidden content as diagnostic signals.

    line 54
  7. Distinctive:

    Keep findings reproducible.

    line 56
  8. Distinctive:

    Run mobile typecheck/tests and both platform exports for mobile/shared changes.

    line 78

≈ “Common” is an estimate: at least 1 in 5 of the 26,980 CLAUDE.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
CLAUDE.md

@AGENTS.md

Claude Code notes

  • Work in the worktree you were started in. Never cd into the main checkout.
  • The desktop app is Kybern GPUIX in apps/gpuix (macOS, Windows, Linux). The Tauri app was removed (#113); the old GPUI client on the gpui branch is archived and must not be ported back.
  • Keep the existing look: reuse apps/gpuix/src/kit before writing new styles, change only what was asked, and screenshot the app window after UI changes.
  • Native GPUIX changes are commits on the Tresnanda/gpuix and Tresnanda/zed forks plus a new pin in apps/gpuix/native/FORK; follow "GPUIX native forks" in AGENTS.md.
  • If you were delegated: never push the "kybern delegation base" commit, wait for CI with blocking commands, and finish with the full report (see "Delegated agents and branches" in AGENTS.md).

Errors people can act on

Every error a feature can show follows "Errors people can act on" in AGENTS.md: a plain headline, one sentence on what to do, a fix button where one exists, and the technical text only behind Show details. Use the shared problem components and describeProblem; never show String(e) or a raw message, and keep the raw-error lint's allowlist empty.

Disk usage and cleanup

Follow the Disk usage and cleanup rules in AGENTS.md for every build and temporary worktree. Check free space before/after heavy work; below 30 GiB free, reclaim task-owned generated files before continuing. Inspect growth at 10 GiB added or 15 GiB in a target directory. Use one checkout-local CARGO_TARGET_DIR and CARGO_INCREMENTAL=0 for disposable builds. Clean unused task-created targets, native DerivedData, dependency installs, and scratch data before finishing; keep deliverables and required validation evidence. This cleanup is part of the task, not an optional follow-up.

Check for active processes/open files first. Preserve running binaries, source changes, worktrees, chat history, backups, simulator app data, signing assets, release archives, and the user's iPadOS 27.0 device support. Do not share the running app's build directory across checkouts or thin Time Machine snapshots as routine cleanup. Report actual free space and explain large retained outputs.

Performance and visual quality

Apply the performance rules in AGENTS.md to every desktop UI change. Before editing a GPUIX rendering hot path, read apps/gpuix/native/README.md (islands, cached views, layout cost) and the GPUIX reports in apps/gpuix/perf/.

  • Preserve the polished kit appearance while bounding rendering work. Reuse stable history, virtual rows, worker queues, paced streaming, and shared visibility handling instead of introducing a parallel implementation.
  • Validate both appearance and interaction: long threads, typing during output, navigation, selection, focus, reading position, and the relevant theme states.
  • Treat fallback text and hidden content as diagnostic signals. Confirm the expected formatted content and reachable controls before claiming success.
  • Keep findings reproducible. Record the workload and measurement boundary; isolated CPU or frame results do not establish whole-app energy savings.

Mobile companion

The Android/iOS app is in apps/mobile; read its README and the Mobile app section of AGENTS.md before editing it. Share protocol, transport, transcript, and composer logic through packages/kybern-client; use the mobile primitives, Ink tokens, and paired SF/Material icon map for native UI. Android uses custom themed components and liquid-gooey-style animation; keep menus opaque and avoid stock Android controls. iOS may use native Liquid Glass.

pnpm start:go --lan targets Expo Go; pnpm start --lan targets an installed development app. EAS preview creates an installable Android APK that runs without Metro. The EAS owner is the personal treshnanda account, never beme-mobile. Release-only native behavior, especially networking, must be checked independently of Expo Go. Keep the iOS scene and Android local-network config plugins until the corresponding platform requirements change.

Mobile has its own app version and EAS build numbers. Daemon, CLI and GPUIX remain one versioned release unit. Pushing main does not publish either one; see RELEASING.md before changing tags or release scripts. Run mobile typecheck/tests and both platform exports for mobile/shared changes. Mobile 0.1.1 supports EAS Update on preview/production channels with native fingerprint compatibility. Native changes need a new build; updates must use the matching EAS environment and runtime. Follow the mobile README for commands.

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.