Legend-State

Use this skill for Legend-State v3 work across core observables, React or React Native integration, persistence, remote sync, and migrations.

Workflow

1. Inspect the project before changing code:

- Read local agent guidance and nearby state conventions. - Check the exact @legendapp/state version and package manager lockfile. - Identify the runtime: vanilla TypeScript, React web, React Native, Expo, SSR, or a mixed app. - Map observable ownership and lifetime: module singleton, request scoped, provider scoped, component local, or synced resource. - Inventory React consumption (useValue, observer, Memo, $React, control-flow components), persistence plugins, sync plugins, and deprecated APIs.

1. Confirm the release channel before using v3 APIs. Version 3 is prerelease in the current snapshot while npm latest is v2. Start with [references/source-map.md](references/source-map.md), then refresh current docs and package metadata when versions matter.
2. For observable creation, reads and writes, computed or linked state, reactivity, arrays, events, batching, and helpers, use [references/core-observables.md](references/core-observables.md).
3. For React, React Compiler compatibility, fine-grained rendering, React Native, hooks, control flow, context, and tracing, use [references/react-integration.md](references/react-integration.md).
4. For local persistence, remote sync, CRUD, retries, transforms, plugins, sync status, and local-first reliability, use [references/persistence-sync.md](references/persistence-sync.md).
5. For v2 migrations, beta drift, architecture, security, performance, SSR, debugging, and tests, use [references/production-migration-testing.md](references/production-migration-testing.md).
6. Match the installed declarations and existing project style. Do not copy v2 examples into v3 code or assume a v3 beta API remains unchanged.

Implementation Judgment

• Prefer useValue(observableOrSelector) for React reads. Treat useSelector and use$ as migration aliases, and do not rely on direct .get() calls inside observer; that pattern conflicts with React Compiler guidance.
• Use observer as an optional hook-count optimization for components with many or conditional useValue calls, not as the primary subscription API.
• Mutate through observable APIs such as set, assign, delete, array methods, and Map or Set methods. Never mutate raw data returned by get() or peek() and expect notifications.
• Use peek() only for deliberately untracked reads. Use get(true) for shallow key or collection tracking when that is the actual dependency.
• Keep computed, linked, and synced getters pure except for the documented data-loading behavior of async or synced getters. Put side effects in actions, observe, onChange, or explicit application services.
• Prefer narrow observable nodes and selectors. Use Memo, $React, For, Show, or Switch only when moving updates to a smaller React boundary materially helps.
• Batch related writes. This matters for listener churn, React renders, and persistence write amplification.
• Choose synced(...) when lazy activation on first read is desired; choose syncObservable(...) when an existing observable should begin syncing immediately.
• Treat persistence as a cache and retry queue, not as authorization or secure secret storage. Keep server authorization, row-level access, validation, and conflict policy explicit.
• For async persistence, gate reads that require hydrated data on syncState(state$).isPersistLoaded. For remote readiness, use isLoaded, error, and pending-change state rather than guessing from data shape.
• Model logout, tenant changes, and auth gates explicitly. A live useValue or other observer can reactivate a lazy synced observable after reset unless the selector or sync configuration is gated.

Verification

Use the repository's existing checks. For meaningful Legend-State work, include the relevant subset:

• Typecheck against the installed v3 beta declarations and exact subpath imports.
• Unit tests for observable writes, computed dependencies, linked setters, batching, cleanup, Map or Set behavior, and raw-data mutation mistakes.
• React tests for render boundaries, useValue, observer, Memo, For, Show, Strict Mode cleanup, Suspense, and React Compiler-sensitive patterns.
• Persistence tests for hydration order, transforms, migrations, retry metadata, out-of-order saves, resets, and storage failures.
• Sync tests for create, update, delete, paging, realtime subscriptions, retries, conflict or last-write rules, auth gating, and pending changes across restart.
• Browser or device smoke tests for offline edits, reload, reconnect, failed writes, logout, tenant switching, and storage upgrades.
• Migration scans for v2 persistence APIs, old computed or proxy helpers, useSelector or use$, direct .get() in render paths, legacy reactive component imports, and renamed return or callback behavior.