documentation

Plugin documentation from marketplace huakun-skills in HuakunShen/skills. Commands read from the repo's manifest — 7 days ago.

Skills (28)

• building-agent-orchestration

  Use when designing, implementing, or reviewing background agents, subagent delegation, parallel research agents, agent teams, task routing, result collection, cancellation, or multi-agent coordination

• dev-loop

  An orchestrator-run development process. The owner drops requests (wishes, bugs, features) into any inbox (Notion, GitHub Projects, a local file); one main agent session records, designs, reviews, slices into parallel lanes, delegates to subagents or other agent CLIs, verifies, merges and reports. D

• injecting-agent-behavior

  Use when adding or reviewing agent commands, command skills, SKILL.md discovery, custom agents, rules, instructions, MCP wiring, context injection, or behavior nudges in coding-agent tools

• model-routing

  Choose models and tools for delegated work, compare them on real tasks, and improve routing from verified outcomes. Use before delegating, when choosing a model or effort level, after rework, or when reviewing model cost and quality.

• porting-agent-plugins-across-harnesses

  Use when porting, designing, or reviewing plugins, skills, commands, tools, MCPs, or agent workflows across OpenCode, Claude Code, Codex, Antigravity (agy), Gemini, Cursor, or multiple coding-agent harnesses

• testing-agent-plugins

  Use when writing, reviewing, or improving tests, evals, smoke tests, integration tests, discovery tests, safety audits, or verification plans for agent plugins, skills, commands, tools, hooks, or orchestration systems

• writing-opencode-plugins

  Use when writing, modifying, debugging, packaging, or reviewing OpenCode plugins, custom tools, plugin hooks, shell-backed tools, npm/local plugin setup, or @opencode-ai/plugin code

• coding-style

  Huakun Shen's personal coding style and engineering standards, distilled from his real repositories (kunkun, refyard, xross-dev, CrossCopy, ListenFlow, space-lens, uniview, model-jetty, logic-foundry, polymarket-quant, localsend-rs, kunkun-services, ...). USE THIS SKILL whenever you write, review, r

• harmonyos-app-patterns

  Working patterns for building real HarmonyOS NEXT / OpenHarmony apps in ArkTS + ArkUI, extracted from a shipping production app (a Clash proxy client with a Go native core). Covers the whole stack: bottom navbar / tab shells, adaptive breakpoints, Navigation + NavPathStack routing, theme colour and

• iroh-development

  Guidance for building peer-to-peer applications with iroh, the Rust-based modular QUIC networking stack from number0 / n0-computer for direct device-to-device connections with NAT traversal, hole punching, relay fallback, and DNS/pkarr-based discovery. Covers the core Endpoint / Router / ProtocolHan

• multi-surface-app

  Design, scaffold, review, or convert a front/back-separated app where one core ships as up to seven thin surfaces: npx CLI, browser web UI, terminal TUI, Tauri desktop app, VS Code extension (webview), DeepSeek Harness (DSH) plugin, and Kunkun custom-view plugin. Stack-agnostic: frontend may be Svel

• native-desktop-styling

  Use when adapting an existing desktop UI to distinct macOS, Windows WinUI/Fluent, and Linux desktop styles, especially when platform themes only change colors and corner radii. Relevant to Tauri/Electron webviews, shared Svelte/React/shadcn components, Tailwind, and equivalent native frontends.

• ratatui-dev

  Build good-looking, feature-rich terminal user interfaces (TUIs) in Rust with ratatui (https://ratatui.rs) and crossterm. Covers layout (Constraint/Flex), every built-in widget (Block, List, Table, Tabs, Gauge, LineGauge, Sparkline, BarChart, Chart, Canvas, Calendar, Paragraph, Scrollbar), third-par

• shadcn-studio

  Use when configuring or using Shadcn Studio Pro/free registries, CLI, MCP, IDE extension, themes, blocks, or adapting Studio React components for Svelte or another framework/language.

• shadcn-svelte-monorepo

  Use when setting up, fixing, or explaining shadcn-svelte in a pnpm/SvelteKit monorepo where generated components live in a shared UI package such as packages/ui and an app such as apps/desktop consumes them through aliases. Trigger for shadcn-svelte components.json, SvelteKit kit.alias, $ui imports,

• ui-component-scout

  Use when finding, comparing, evaluating, or integrating beautiful UI components, shadcn registries, animated React primitives, AI-chat interfaces, distinctive cards, tables, graph visuals, media players, and their Vue/Svelte/Solid ports. Provides scenario-based discovery, an offline component yellow

• webview-native-look

  Use when a web UI hosted in an embedded WebView (WKWebView in an AppKit/Swift shell, Tauri, Electron) must look and behave like a native macOS app - real vibrancy / Liquid Glass materials, system accent colour, light/dark and OS-generation-aware styling, UI zoom, no "web page" tells (selection, scro

• cloudflare-ai-search-manual-sync

  Create, manage, sync, and maintain a Cloudflare AI Search instance with manual document upload via the items API. Unlike the R2/web-crawl auto-index approach, this covers the admin API pattern — list, diff, upload, delete, and sync. Use when the user wants to build a custom content pipeline for AI S

• cloudflare-fumadocs-rag-agent

  Add a RAG AI chat agent to a Fumadocs documentation site using Cloudflare AI Search. Covers provisioning, wrangler config, Hono API, content sync script, and the floating chat UI. Use when the user asks to add AI search, chat, or RAG to a Fumadocs project. This is the Cloudflare-specific version.

• cloudflare-monorepo

  Multi-Cloudflare-Worker monorepo pattern using pnpm workspace + turbo. Covers apps/api (Hono + WorkerEntrypoint RPC), apps/web (SvelteKit or TanStack Start), packages/db (shared Drizzle/D1 schema), and typed service-binding RPC between workers. Use when scaffolding, modifying, or reviewing a CF mono

• building-codebase-curriculum

  Use when someone wants to learn, re-learn, or teach a codebase that already works but whose implementation they do not understand — a project written mostly by an AI agent whose author only knows the direction and not the mechanics, onboarding onto a large unfamiliar repo, or a request to turn a fin

• deepwiki-badge

  Generate a DeepWiki badge (Markdown or HTML) from a repo URL, or from the current git remote if run inside a repository, and optionally insert it into README.md.

• journal

  Reviews recent changes and decisions, and documents them in a dated journal file.

• sync-wiki

  Sync .repowiki documentation based on recent changes. Reads .journal entries, .reports experiment results, and git diff to update relevant wiki pages. Use when user mentions "sync wiki", "update repowiki", or after completing significant work.

• ai-web-computer-use

  Use when an agent must operate the visible, logged-in web UI of ChatGPT, Gemini, or NotebookLM with Computer Use alone to ask questions, continue or start conversations, generate media, or download files.

• external-mobile-sdks

  Use whenever work needs an Android or HarmonyOS SDK, emulator/AVD, adb, or hdc on this Mac, or when Android Studio/DevEco reports a missing SDK. Resolves the real SDK locations through the external-disk symlink layer at /Volumes/Portable2TB/MobileDev instead of assuming default internal paths.

• slidev-narrated-video

  Turn a Slidev Markdown/Vue deck into a locally rendered, audio-synchronized MP4 by compiling a validated per-(page, click) narration manifest, capturing deterministic 1920x1080 slide states with Playwright, assembling them with Remotion, and verifying the result. Use when a developer deck needs both

• strip-ai-coauthors

  Remove AI / agent co-author and attribution trailers (Co-Authored-By: Claude, Sisyphus, [email protected], "Ultraworked with", "Generated with Claude", 🤖 lines) from git commit messages by rewriting history. Use when a user wants to clean up, strip, or scrub co-author / bot / AI-attribution tra

Similar plugins