AGENTS.md

Workspace

core/ owns the engine, CLI, daemon, SDK, native helpers, and releases. Run Cargo there; apply relevant core/CODING_STANDARDS.md rules when changing core code.
dashboard/ owns Ployz Cloud: web, worker, durable workflows, and self-host packaging. Run pnpm there.
Before designing Engine behavior, including Cloud code that drives it, read the bets in core/DESIGN.md. Name each bet your change touches in the PR body, and the choice it changed. If you cross a red flag, argue it there. Dashboard UI follows dashboard/DESIGN.md.
Use the terms defined in core/CONTEXT.md (Engine) and dashboard/CONTEXT.md (Cloud) rather than synonyms. Work across the Engine–Cloud boundary reads both.
docs/user/ describes shipped behavior from the user's point of view. Check it before deciding how something works today, and update it when user-visible behavior changes.
Prove changes on the real artifact: dashboard UI with the verify-dashboard skill (add SERVERS=1 for runtime UI), and anything needing real Machines (daemon, Docker, WireGuard, ZFS, moves) with the verify-server skill, before reaching for cloud VMs.

GitHub CI is the PR gate. Locally, run only focused checks: typecheck, lint, fmt, and the tests for the code you changed. Don't run a full suite (pnpm pr:check, pnpm test, cargo nextest run, workspace clippy) locally, except to reproduce a CI failure.
Opening, reviewing, verifying and landing PRs: follow docs/agents/landing.md.

Draw a diagram where one picture compresses many words: how systems relate, flows, forks. Show UI with screenshots or prototypes, never ASCII.