agentleFS
young1theyoung1the/pets-driven/AGENTS.md

pets-driven

AGENTS.md1 starsChanged 3 months agoMITView on GitHub

Pets-Driven is a Windows desktop app (Tauri + React) where a pet represents one bound coding-agent execution, plus the landing site and shared packages around it. Start with CONTEXT.md for the domain language and ARCHITECTURE.md for how the modules connect. pet-engine depends on nothing in the workspace and must stay that way. Full graph, runtime event flow, and a ripple table: ARCHITECTURE.md. The Rust side is a Cargo workspace (root Cargo.toml): the desktop Tauri crate plus crates/pets-driven-core (persisted Pet +…

  • 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 31,774 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 repository3 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, 1 is distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    All code comments, JSDoc, and in-repo documentation (Markdown, ADRs, READMEs) must be written in English.

    line 7
  2. Very specific:

    Every cross-package import goes through the target package's declared exports entry (@pets-driven/<pkg>/...), never a deep relative path …

    line 81

≈ “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

AGENTS.md

Pets-Driven is a Windows desktop app (Tauri + React) where a pet represents one bound coding-agent execution, plus the landing site and shared packages around it. Start with CONTEXT.md for the domain language and ARCHITECTURE.md for how the modules connect.

Language

  • All code comments, JSDoc, and in-repo documentation (Markdown, ADRs, READMEs) must be written in English.
  • Git commit messages and pull request descriptions must also be written in English.
  • This applies to source files, type definitions, test descriptions, and any docs committed to the repository.
  • Conversations and chat with the user may continue in Korean; only artifacts written to disk or git history are English-only.

Commands

pnpm dev              # desktop app (Tauri)
pnpm dev:playground   # simulation playground in the browser
pnpm dev:design       # design-system component gallery
pnpm test             # every package's suite + scripts/*.test.mjs
pnpm typecheck        # tsc across the workspace
pnpm check            # biome lint + format check
pnpm sync-pets        # copy repo-root pets/ into each app's public assets
pnpm encode-demos     # re-derive docs/assets/*.mp4 + poster from the GIFs (needs ffmpeg)
pnpm test:e2e         # Playwright, desktop only
pnpm test:path-hook    # the installer's PATH edit, against a scratch registry key
pnpm release:bump <major|minor|patch|X.Y.Z>  # one version across every manifest, committed and tagged

Modules

PathWhat it ownsDocs
apps/desktopTauri shell, pet windows, agent event ingressapps/desktop/AGENTS.md
crates/pets-driven-coreAuthoritative Pet + Registered Working Directory state behind a repository seamcrates/AGENTS.md
crates/pets-driven-fsShared on-disk state.v1.json repository with a cross-process file lockcrates/AGENTS.md
crates/pets-driven-protocolHook-forwarding wire contract (routes, synthesized event)crates/AGENTS.md
crates/pets-driven-gitGit worktree behavior the app and the CLI share (where a worktree goes, what git is asked)crates/AGENTS.md
crates/pets-driven-cliThe pdd CLI: direct state ops + live hook forwardingcrates/AGENTS.md
apps/webLanding site + Remotion demo videoapps/web/AGENTS.md
packages/pet-engineECS simulation, personalities, sprite statepackages/pet-engine/AGENTS.md
packages/design-systemTokens and shared React componentspackages/design-system/AGENTS.md
packages/i18nLocales, i18next setup, translation catalogpackages/i18n/AGENTS.md
plugins/pets-drivenShared Claude Code + Codex plugin: hooks, Claude commands, Codex skills—
pets/Canonical built-in pet definitions and spritesheetspets/README.md
scripts/Asset sync, version bump, install test—
graph TD
  desktop[apps/desktop] --> engine[packages/pet-engine]
  desktop --> ds[packages/design-system]
  desktop --> i18n[packages/i18n]
  desktop --> core[crates/pets-driven-core]
  desktop --> fs[crates/pets-driven-fs]
  desktop --> git[crates/pets-driven-git]
  cli[crates/pets-driven-cli] --> core
  cli --> fs
  cli --> git
  git --> core
  web[apps/web] --> engine
  web --> ds
  web --> i18n
  ds -.->|dev only| engine

pet-engine depends on nothing in the workspace and must stay that way. Full graph, runtime event flow, and a ripple table: ARCHITECTURE.md.

The Rust side is a Cargo workspace (root Cargo.toml): the desktop Tauri crate plus crates/pets-driven-core (persisted Pet + Registered Working Directory behavior behind a StateRepository seam; depends only on serde, serde_json, thiserror), crates/pets-driven-fs (the shared FileStateRepository — the on-disk state.v1.json plus an fslock cross-process lock), crates/pets-driven-protocol (the hook-forwarding wire contract), crates/pets-driven-git (the git worktree behavior the desktop and the CLI share), and crates/pets-driven-cli (the pdd binary).

The desktop and the pdd CLI both write state directly through the core over the same pets-driven-fs repository — the same file, resolved identically (dirs::data_dir()/com.petsdriven.desktop/state.v1.json, overridable with PETS_DRIVEN_STATE_PATH), and serialised by the same cross-process lock. So pdd works whether or not the desktop is running and cannot race it. A write is a durable state change; a hook event is not — it is a transient signal a running pet reacts to, so pdd forward goes to the live app over the loopback ingress and no-ops when the app is down. The desktop watches the state file and reloads the webview when the CLI writes it.

Non-obvious rules

  • Gotcha: pets/ at the repo root is the source of truth for built-in pets. pnpm sync-pets copies it into each app's git-ignored public assets directory. Editing the copied output looks like it works until the next sync silently reverts it.
  • Note: the pet window is a fixed 192x268 always-on-top overlay. The OS window never shrinks, so the projection must center that fixed window rather than the scaled sprite frame — otherwise small pets sink and the collision capsule clips.
  • Why the engine is a separate package: the simulation must stay deterministic and headless so it can be tested without a window. Keep Math.random(), Date.now(), and DOM access out of it.
  • Don't add a system without registering it in the pet-engine phase pipeline (packages/pet-engine/src/core/phases.ts); an unregistered system silently never runs.
  • Caveat: pnpm test is the full workspace suite and is slow. For a change scoped to one package, run that package's own suite with pnpm --filter <pkg> test. A pure token hex edit needs neither the desktop suite nor a typecheck.

Conventions

  • Each module keeps its context in its own AGENTS.md, with a one-line CLAUDE.md beside it that reads See [AGENTS.md](./AGENTS.md). — content lives in exactly one file.
  • Every cross-package import goes through the target package's declared exports entry (@pets-driven/<pkg>/...), never a deep relative path across a package boundary.

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.