lobehub/lobe-chat
30 skills20.6k installs2M starsGitHub
Install
npx skills add https://github.com/lobehub/lobe-chatSkills in this repo
1Drizzledrizzle is a schema style guide skill in lobehub/lobe-chat for Drizzle ORM work on LobeHub’s Postgres database. It applies when editing pgTable definitions, indexes, joins, inferred types, db.select and db.query calls, foreign keys, junction tables, and repository code under packages/database. The skill requires sibling tests for new models and repositories in __tests__ with the getTestDB integration pattern documented in the testing skill. Developers reach for drizzle when agents touch LobeChat database files so schemas, queries, and user-isolation tests stay consistent with existing conventions. It prevents one-off SQL styles that diverge from LobeHub’s established Drizzle patterns and BM25 server guards.1.7kinstalls2Typescripttypescript is LobeHub's TypeScript code style guide packaged as an agent skill for editing TS, TSX, and MTS across skills and MCP servers. It directs developers to avoid explicit annotations when inference suffices, eliminate implicit any, prefer Record<PropertyKey, unknown> over object or any, and choose interface for object shapes while reserving type for unions and intersections. The guide recommends as const satisfies over plain as const and @ts-expect-error over @ts-ignore for intentional suppressions. Developers invoke typescript when fixing types, choosing interface versus type, handling async flow, or reviewing agent-shipped TypeScript. The skill is user-invocable false, meaning agents apply it automatically during relevant edits rather than on direct user command.1.6kinstalls3Reactreact is a Skillselion skill from lobehub/lobe-chat that documents @lobehub/ui layout primitives for React applications. The guide covers Flexbox, the most common layout component mirroring CSS display flex, plus Center for alignment tasks. Developers learn default vertical stacks, horizontal rows, and prop controls including horizontal, flex, gap, and align values such as center and flex-start. Reach for this skill when building Lobe Chat or other React apps that standardize on @lobehub/ui instead of hand-rolling flex CSS for every screen.1.6kinstalls4Zustandzustand is a LobeHub skill from lobehub/lobe-chat that documents Zustand store conventions for AI chat and agent UIs under src/store. It defines public, internal, and dispatch action layers, an optimistic update pattern, slice composition via flattenActions, and class-based action migration from StateCreator object slices to XxxActionImpl classes. Triggers include useChatStore, useUserStore, useGlobalStore, createStore, StoreSetter, internal_dispatch, messagesMap and topicsMap reducers, and debugging stale store reads. Developers reach for zustand when adding createXxxSlice functions, writing internal_* or internal_dispatch* actions, refactoring slices into class actions, or fixing selector staleness in LobeChat. The skill keeps multi-slice chat state consistent across topics and messages while separating public API actions from internal dispatch plumbing required for optimistic UI updates in conversational interfaces.1.6kinstalls5I18ni18n is the LobeHub internationalization skill for lobe-chat using react-i18next with default language en-US. Developers edit only files in src/locales/default/ and never edit JSON files in locales/ directly; run pnpm i18n to generate translations or manually update zh-CN and en-US for dev preview. Keys use flat dot notation such as alert.cloud.action rather than nested objects. The skill triggers on user-facing strings, locale keys, namespaces, useTranslation, t(), interpolation, hardcoded UI copy, or pnpm i18n commands. Reach for i18n when adding or refactoring Lobe Chat UI text to avoid key conflicts and preserve locale consistency across zh-CN and en-US previews.1.6kinstalls6Project Overviewproject-overview is a user-invocable-false onboarding skill from lobehub/lobe-chat for the LobeHub open-source monorepo (github.com/lobehub/lobehub, package @lobehub/lobeh). It provides a curated map of key locations across apps, packages, src/store, and route groups—not an exhaustive tree, since directories grow over time. Developers reach for project-overview when adding features, creating new agents, extending the workspace, or understanding business stub layers in the modern-design AI Agent Workspace formerly known as LobeChat. The skill directs contributors to run ls against real directories for current listings. Use it as the first stop before navigating the monorepo for frontend, backend, or agent changes.1.6kinstalls7Desktopdesktop is an agent skill from lobehub/lobe-chat for developers adding native desktop capabilities to an AI-powered chat or agent application built on Electron. The architecture splits main-process Controllers handling IPC from renderer-process Service Layers and Store Actions managing UI state. Implementation steps start with creating a Controller IPC handler, wiring system APIs for filesystem and network access, and connecting renderer services that invoke native notifications and OS integrations. Reach for desktop when extending Lobe Chat or a fork with file pickers, local file reads, desktop notifications, or other main-process privileges unavailable in pure web builds. The guide uses TypeScript across main and renderer boundaries with explicit IPC contracts.1.5kinstalls8Add Setting Envadd-setting-env is a Lobe Chat repository skill with disable-model-invocation set true, guiding maintainers through adding server-side environment variables that control default user settings. It follows a priority chain: User Custom beats Server Env Var beats Hardcoded Default. Steps include creating src/envs/<domain>.ts with createEnv from @t3-oss/env-nextjs, Zod coercion and min/max validation, wiring env reads into setting defaults, and documenting the variable. Developers reach for add-setting-env when shipping a new configurable Lobe Chat preference without editing hardcoded defaults scattered through the codebase.1.5kinstalls9Add Provider Docadd-provider-doc is a LobeChat maintainer workflow skill with disable-model-invocation enabled and an argument-hint of [provider-name]. It walks through five deliverables: English and Chinese usage pages at docs/usage/providers/{provider-name}.mdx and {provider-name}.zh-CN.mdx, matching environment-variable documentation in both locales, Docker configuration updates, .env.example changes, and provider image resources. The skill standardizes how new OpenAI-compatible or vendor-specific providers appear in LobeChat docs so Docker deployers and self-hosters can configure keys and endpoints consistently. Developers reach for add-provider-doc when landing a new provider PR and need EN+CN docs, env references, and container config updated in one pass rather than hunting scattered doc patterns.1.5kinstalls10Modalmodal is a lobehub/lobe-chat skill marked user-invocable false that documents LobeHub imperative modal conventions on the base-ui stack instead of antd Modal. New code should import createModal, confirmModal, and ModalHost from @lobehub/ui/base-ui, with useModalContext inside modal content components. Body content passes via content or children, resolved as content ?? children at runtime. A global ModalHost is required because base-ui createModal renders through the host. The skill covers creating, migrating, and wiring confirm flows and popups consistently. Developers reach for modal when adding LobeChat dialogs or replacing legacy antd modals with headless base-ui primitives.1.5kinstalls11Testingtesting is the Agent Runtime E2E Testing Guide from lobe-chat. It follows a Minimal Mock Principle: only three external dependencies are mocked—database via PGLite from `@lobechat/database/test-utils`, Redis via InMemoryAgentStateManager, and stream events via InMemoryStreamEventManager. Model-bank and other internals run with real configuration rather than stubs. Developers reach for this skill when writing E2E tests for AI agent runtimes that must exercise realistic paths without standing up production databases or Redis. The approach prioritizes faithful runtime behavior over heavy isolation.1.5kinstalls12Hotkeyhotkey is a LobeChat contributor skill that walks through the project's keyboard shortcut system step by step. It covers updating HotkeyEnum in src/types/hotkey.ts, registering defaults in src/const/hotkeys.ts via HOTKEYS_REGISTRATION, wiring combineKeys from @lobehub/ui, and connecting useHotkeyById hooks with tooltip labels. Developers reach for hotkey when extending LobeChat with new shortcuts, fixing scope conflicts, or maintaining Cmd versus Ctrl platform differences. The skill documents group assignments via HotkeyGroupEnum and the full registration shape so new hotkeys behave consistently across the chat interface without breaking existing bindings.1.5kinstalls13Linearlinear is a LobeHub issue-management skill for LOBE-xxx Linear tickets that enforces two jobs on every fix PR: a Fixes LOBE-xxx keyword in the PR body so Linear auto-links and auto-closes on merge, plus a completion comment on the Linear issue for humans. Before running workflows, the skill instructs agents to search for linear MCP tools and treat missing tools as not installed. Developers reach for linear when Claude Code or similar agents create PRs that resolve tracked issues and need machine-readable GitHub magic keywords paired with readable stakeholder updates. It also covers retrieving issues, updating status, and navigating sub-issue trees during active development cycles.1.5kinstalls14Db Migrationsdb-migrations is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.12installs15Store Data Structuresstore-data-structures is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.12installs16Agent Tracingagent-tracing is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.11installs17Chat Sdkchat-sdk is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.11installs18Clicli is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.11installs19Prpr is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.11installs20Response Complianceresponse-compliance is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.11installs21Spa Routesspa-routes is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.11installs22Trpc Routertrpc-router is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted development.11installs23Upstash Workflowupstash-workflow is a Claude Code skill for automation & workflows. It helps solo builders move faster with AI-assisted coding.11installs24Version Releaseversion-release is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.11installs25Agent Runtime Hooksagent-runtime-hooks is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.10installs26Agent Signalagent-signal is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.10installs27Builtin Toolbuiltin-tool is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.10installs28Debug Packagedebug-package is a Claude Code skill for debugging. It helps solo builders move faster with AI-assisted coding.10installs29Docs Changelogdocs-changelog is a Claude Code skill for git & pull requests. It helps solo builders move faster with AI-assisted coding.10installs30Heterogeneous Agentheterogeneous-agent is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.10installs