LobeHub

Use for Zustand stores: list/detail splits, state type sources, slices, actions, reducers, selectors, optimistic updates and class-action composition.

Use for public LobeHub website documentation under docs/**, including self-hosting, deployment, migration, and configuration guides. Excludes product changelogs and source-code API docs.

Version release workflow — release process and GitHub Release notes (not docs/changelog pages).

Audit a page / surface against the Designing Interfaces pattern language + the ux skill checklists, then land findings. Three layers — static (code), visual (screenshots), dynamic (automated user journey + perf). Use to run a repeatable, standards-based UX review of one screen.

Use for product design principles and UX checks when designing or implementing user-facing flows.

Use for Upstash Workflow/QStash handlers, triggers, durable steps and async fan-out.

Use for TypeScript style and type safety when editing TS/TSX/MTS, including imports, async code and error suppression.

Use for server TRPC routers, procedures and API endpoints in apps/server/src/routers.

Manually runs the live LobeHub official-provider compatibility matrix across Claude Code, Codex, Grok Build, Kimi Code, Pi, and TRAE when explicitly invoked by the user.

Use for Vitest tests, mocks, coverage and failing-test diagnosis.

Explicit compatibility entry for Zustand data-shape guidance.

Use for standalone micro apps: React Router SSR/SSG, Cloud overlays, Workers, CDN/R2, SEO/OG, gateway routes, previews, Docker integration and SSR bundle isolation.

Use for SPA routes and page segments: Web/desktop/mobile/popup routers, redirects and src/routes-to-features boundaries.

Audit .agents/skills SKILL.md files. Use for recurring checks of duplicate, overlapping, stale, inconsistent, or broken skills and merge/delete candidates.

Run and diagnose existing Harbor evaluations against local production LobeHub or LobeHub Cloud. Use for eval infrastructure, target preflight, lh CLI injection, Harbor smoke or job runs, resume, and failure triage. Excludes authoring Harbor tasks and product acceptance.

Use for OpenResponses/Response API compliance tests, endpoint schema failures and test runs.

Use for TSX components, UI libraries, styling, state locality, layout, render performance and memoization.

Use for the OSS repository map, code-layer lookup, apps/packages/src ownership and business stubs.

Use for new surfaces, redesigns and vague product asks: establish user tasks and business meaning. Built-screen audits and visual polish have separate skills.

Create a PR for the current branch (targets `canary` by default), including splitting one cross-layer branch into ordered stacked PRs so a lower layer (db / shared package / server TRPC) merges before its callers (desktop / CLI / UI). Use when the user asks to create / submit a PR, or to split a branch because clients call a server contract that isn't on the trunk yet. Triggers on 'pr', 'create pr', 'submit pr', 'open a PR', 'pull request', 'split this PR', 'stacked PR', 'backend should merge first', '提 PR', '提个 PR', '新建 PR', '拆 PR', '后端先合', '分层合并'.

Use for model knowledgeCutoff, family and generation metadata: onboarding, corrections, research and bulk backfills.

Use for modals, dialogs and confirmations with createModal, confirmModal, ModalHost or base-ui modal APIs.

Use for application prompts, generateObject/generateText, model selection and generation tracing. Excludes provider adapters and agent snapshots.

Use for Linear issues, LOBE-xxx links, status updates, completion comments and sub-issue trees.

Use for user-facing strings and react-i18next locale keys, namespaces, interpolation, translations or bun run i18n.

Use for keyboard shortcuts, registration, key combinations, scope, conflicts and shortcut tooltips.

Use for Claude Code/Codex external-agent adapters, IPC, event mapping, sessions, persistence and tool-call chains.

Use for product search: FtsSearchRepo, pg_search/pg_like/Elasticsearch, mapping migrations, projections, Outbox sync, reindexing and performance. Excludes agent web search.

Use for Drizzle schemas and queries: tables, indexes, relations, joins and inferred types. Rollout belongs to db-migrations.

Use for EN/ZH website product changelogs in docs/changelog/*.mdx. Excludes GitHub Release notes.

Electron desktop development guide — IPC handlers, controllers, preload scripts, window/menu management.

Use for interactive HTML prototypes (交互原型) with the real LobeHub UI stack. Excludes production UI implementation.

Use for independent PR, diff or branch review; not explanations. Full multi-agent review requires an explicit deep-review request.

Use for debug() logging, lobe-* namespaces, DEBUG/localStorage.debug configuration and log formatting.

Use for intermittent UI, stale state, ordering, navigation or browser-only bugs. Extends acceptance with boundary tracing and regression diagnosis.

Use for Drizzle migration rollout, online indexes, backfills, idempotent SQL and migration regeneration or rebase conflicts.

Use for client APIs, SWR hooks, cache invalidation, async errors, useEffect migration, home first paint and persistent caches.

Use for splitting heavy domains into capabilities (原子组件) across page, portal, share or micro-app hosts, sinking state (状态下沉) and replacing mode/readOnly flags. Simple component splits: react.

LobeHub CLI (@lobehub/cli) development guide — commands, subcommands, architecture.

Use for stale-worktree, Git registration and branch audits. Managed removal uses the repo lifecycle.

Use for multi-platform bots built with chat SDK: webhooks, mentions, slash commands, cards, modals and streaming.

Use for LobeHub builtin agent tools: manifests, executors, runtimes, inspectors, renders, streaming and intervention.

Use for the Agent Work output registry: registration, deduplication, cards, skill providers, CLI scanners and Work types.

Use for agent traces by operation ID (拉线上 tracing): failed tools, arguments/results, available tools, execution location, LLM calls and context.

Use for real bot-channel acceptance in Discord, Slack, Telegram, WeChat/Weixin, Lark/Feishu, QQ or iMessage on macOS. Extends acceptance with native chat apps.

Use for Agent Signal sources, actions, policies, middleware, workflow handoff and deduplication.

Use for agent lifecycle hooks, tool mocks, intervention, sub-agent calls and context compression.

Add server-side environment variables that control default values for user settings.

Add an AI model provider to LobeHub, including runtime integration, model cards, configuration, branding, documentation, and acceptance. Also use for provider documentation updates; adding models to an existing provider does not require a new provider integration.

End-to-end verification and self-evidence for a delivery in any repository, with or without a preconfigured verify plan. Discover an existing plan when one was handed to this run; otherwise author checks and publish a standalone acceptance. Pick the proving surface (CLI / web / desktop / iOS Simulator), drive the real product, capture visually confirmed evidence, and publish a round with the lh CLI. Triggers on 'verify the task', 'collect evidence', 'prove it works', 'upload evidence', 'verify plan', 'requiredEvidence', 'local test', 'manual test', 'test report', 'test with cli', 'test in electron', 'test desktop', or any local end-to-end verification task. Needs no ambient ids, and never depends on running inside a LobeHub conversation.