@openlark

zx wraps child_process to let you write shell scripts in javascript/typescript with auto-escaped arguments and sensible defaults. covers command execution, piping, error handling, and real-time streaming through a clean tagged template api.

mediapipe is google's on-device ml framework for vision, text, audio, and llm inference across android, ios, web, and desktop. covers pre-built tasks, custom pipeline graphs, model customization, and cross-platform apis.

full-link-data-analysis provides a seven-layer workflow for translating business questions into structured reports via python code generation, data-aware method routing, and quality gates. targets analysts and decision-makers across multiple problem types.

dotenv loads environment variables from .env files into node.js process.env with minimal setup. covers basic loading, multi-environment workflows, encryption via dotenvx, and typescript integration.

imapflow wraps a modern node.js imap client with async/await, auto-extension detection, mailbox locking, and provider-specific connection helpers. covers fetching, searching, idle push notifications, and gmail-specific label operations.

multi-agent-communication provides two core primitives (sessions_spawn and sessions_send) for launching child agents and inter-process messaging. covers task decomposition, parallel processing, and multi-round agent negotiation with explicit security constraints and execution flow documentation.

clack-prompts provides a minimal, tree-shakeable toolkit for building interactive node.js cli apps with pre-styled components for text input, selection, spinners, and task runners. covers cancellation guards, sequential prompts, validation, and streaming output.

undici is a node.js http client offering request(), fetch(), stream(), and pipeline() apis with agent-based connection pooling, proxy support, mock testing, and cache interceptors. notably faster than axios for concurrent workloads.

generates social media cover images by templating html based on user-supplied titles and platform dimensions, then converts to png via headless browser. supports 11 major platforms with preset aspect ratios and color scheme suggestions.

data-analysis-report-generator ingests excel/csv files, auto-detects schema (dimensions, metrics, timelines), runs parallel analysis, and renders self-contained html reports with interactive echarts in 11 professional publication styles.

miniprogram-automator provides node.js bindings for controlling wechat mini program ui in devtools, covering page navigation, element selection, data injection, screenshots, and event listening for e2e and regression testing.

lyra-prompt-architect guides users through a four-phase dialogue-to-refinement process for engineering high-performance prompts, selecting reasoning frameworks (cot/tot/got/aot) and applying metacognitive verification techniques for complex tasks.

open-ragflow covers docker deployment, configuration of llm providers and document engines, cli-based knowledge base management, and runtime troubleshooting for self-hosted rag infrastructure.

search-synthesis-expert decomposes complex research tasks into searchable sub-problems, executes multi-source browser automation via playwright, and synthesizes findings with source attribution and credibility assessment.

xiaohongshu-longpost-auto orchestrates end-to-end publishing of long-form content on xiaohongshu's creator platform, from login detection through content segmentation, formatting, tagging, and publication, with manual qr-code login as the only non-automated step.

locateanything provides inference and fine-tuning for nvidia's 3b vision-language grounding model, supporting object detection, text grounding, gui element location, and ocr via parallel box decoding. covers api, data prep, training, and evaluation across coco, lvis, and custom benchmarks.

minimatch converts glob patterns into javascript regex for file path matching, supporting extended globs, brace expansion, and globstar. security-critical: never pass untrusted input as patterns due to redos risk.

guardian-auto-healer runs a cron-scheduled watchdog that checks openclaw gateway health, memory, and disk space every 5 minutes, triggering automatic restarts when thresholds are exceeded.

gray-matter extracts and stringifies YAML/JSON/TOML front-matter from markdown files via command-line or programmatic API, with auto-detection of delimiter formats and custom engine support.

token-cost-optimization provides a three-tier framework for reducing api costs through prompt compression, conversation caching, and model routing, with a calculator tool and phased implementation roadmap.

miniprogram-ci wraps wechat devtools compilation for automated code upload, preview generation, npm builds, and cloud deployments via node.js or cli. covers mini program, mini game, and plugin types with concurrent robot support.

skill-workflow-orchestrator chains multiple sub-skills into automated pipelines with sequential execution, conditional branching, and basic error handling. parses natural language into structured skill sequences and supports templated workflows for common patterns.

storyboard-master converts creative briefs into shot-by-shot storyboard tables with standardized fields for size, camera movement, dialogue, duration, lighting and sound. serves video production pre-production workflows.

ai-prompt-optimization-expert applies the crisp framework to diagnose and refactor user prompts for clarity and precision. outputs diagnostic scoring and revised versions with explicit improvement notes.

text-to-svg converts natural language into valid svg code, covering flat, gradient, stroke, and tech styles with clear output contracts and reference palettes. lacks specificity on error recovery and input validation constraints.

perryts compiles typescript to native binaries via swc and llvm, with declarative ui frameworks (perry/ui for native controls, perry/tui for terminals), multi-threading primitives, and cross-platform deployment targets. covers cli, project config, standard library bindings, and plugin architecture.

web-ui-builder generates react frontends from ui descriptions using a fixed stack of react 19, vite 8, tailwind css 4, and lucide-react. outputs component files, project structure, and startup commands.

fish-speech wraps the s2-pro dual-ar tts model with vllm and sglang server options, supporting voice cloning, emotion tags, and 80+ languages at 44.1 khz output. deployment-focused rather than intent-driven.

wechat-mp-article-publisher automates article creation and publishing on wechat official accounts via browser automation, handling login prompts, content population, ai cover generation, and multi-factor verification flows.

model-prompt-generator converts chinese clothing descriptions into english prompts for ai image generation, covering model appearance, outfit coordination, scene lighting, and photography style within 700 character bounds.

product-prompt-generator constructs english prompts for ai image tools from product metadata. workflow parses product info, generates styled prompts with commercial photography defaults, optimizes keywords, and outputs under 500 characters.

natural language data query assistant generating sql, executing across multiple databases, and rendering results as markdown summaries or interactive html charts. supports multi-turn refinement and file-based data sources.

tech-translator automates chinese-english technical documentation translation with format-aware handling for markdown, json, yaml, html, and 12+ other formats. preserves structure, terminology consistency, and protected content (code, apis, placeholders) across batched files.

wechat-article-generation-expert generates 1600+ word official account articles from topics, audience, and style preferences. includes title ideation, structural planning, and multimedia annotation but lacks explicit failure handling and decision gates.

agent-monitor detects subagents unresponsive for over 5 minutes and sends activation messages to resume work. intended for long-running task oversight and automated health checks across agent fleets.

ai-prompt-optimization provides diagnosis, templating, few-shot generation, and chain-of-thought design for refining ai prompts across multiple platforms. covers structured workflows and diagnostic checklists but lacks concrete decision logic and failure recovery.

lego-pixel-art converts images to brick layouts with material lists via a frontend tool. accepts user images, applies lego color matching, and outputs building guides with purchase requirements across 10-200 stud dimensions.

prd-writing-expert scaffolds structured product requirements documents using moscow prioritization, user stories, and success metrics. covers problem definition through document assembly but lacks procedural depth and failure handling.

toutiao-graphic-publisher automates article publishing to toutiao through browser automation, including intelligent heading detection, tag generation from content analysis, and formatting optimization. requires manual qr code login once, then handles content filling and publication automatically.

skill-distiller is a meta-skill for converting fuzzy domain knowledge into repeatable operational processes through problem definition, theoretical grounding, step-by-step structuring, and tool provisioning.

xhs-suggest-keywords automates xiaohongshu search suggestion collection via playwright. collects auto-suggest dropdowns for seed keywords to support seo and content strategy workflows.

competitive-dimensions-analysis guides structured competitor research across feature matrices, positioning analysis, and swot frameworks. procedure is sequential but relies on manual data gathering without automation or vendor integration.

task-orchestrator decomposes complex goals into executable task trees, matches agents to work items, and monitors progress. lacks concrete failure recovery and decision point validation.

xiaohongshu-image-auto orchestrates end-to-end publishing on xiaohongshu's creator platform given user-supplied title and body text, handling login detection, ai image generation, tag activation, and content declaration through browser automation.

agent-base helps developers create custom http.Agent subclasses for proxying, socket routing, and connection handling. includes quickstart example and references to mature proxy implementations.

deep-article-analysis offers a multi-framework approach to article interpretation using 10+ thinking models (scqa, 5w2h, first principles, six thinking hats, etc). workflow spans quick scan, structure extraction, in-depth analysis, and insight extraction with prescribed output format.

d3js covers low-level data visualization primitives across 30+ modules, including selections, scales, shapes, transitions, and framework integration. primary use case is custom svg/canvas rendering when off-the-shelf charts are insufficient.

model-replacement-prompt-generator builds prompts for swapping characters and scenes in images while locking clothing and footwear. analyzes uploaded images to extract gender, pose, apparel details, and season, then generates constrained prompts under 300 characters.

storyboard-prompt-generator converts anime storyboard scripts into four templated prompt types (character, scene, video, standard) with bilingual output and structured vocabulary, but lacks step-by-step procedural detail and failure handling.

multi-agent-collaboration-communication provides conceptual frameworks for designing distributed agent systems, covering architecture patterns, task decomposition, and communication topology selection without prescriptive implementation guidance.

xiaohongshu-insight aggregates 2000+ daily viral posts across categories using engagement and growth thresholds, offering creators trend analysis and content reference data. lacks clarity on actual invocation, error handling, and data freshness guarantees.

product-pose-prompt-generator assembles image generation prompts for product-holding character scenes by extracting visual details from user descriptions and reference images, then synthesizing them into coherent english text within 300 words.

google-web-fonts documents how to construct google fonts api urls for web typography, covering font selection, weights, styles, subsets, and visual effects. lacks guidance on api limits, caching, fallbacks, or failure modes.

novel-writing-techniques catalogs six literary methods (dual narrative, character biography, foreshadowing, poetic language, moral gray characters, intellectual duels) with routing logic to modular reference files. lacks concrete procedures and failure handling.

playcanvas covers web-first 3d graphics with engine api, editor, react wrapper, and web components, but lacks actionable procedures for common tasks like scene setup or asset publishing.

pi-tui is a component-based terminal ui framework offering differential rendering, ime support, and synchronized output to eliminate flicker in interactive cli applications.

idea-validator outlines a framework for startup evaluation across market, competition, and technology dimensions, but lacks concrete procedural steps, decision logic, and failure mode handling needed for reliable execution.

wmp-to-uniapp converts native weChat mini programs to cross-platform uni-app projects via a 7-step flow covering config migration, page/component conversion, and api remapping. structured but incomplete guidance with vague handling of edge cases.

tanstack-libraries covers a broad ecosystem of headless, type-safe web tooling across data management, routing, tables, forms, and performance. references external docs rather than providing actionable procedures.

batch-content-factory describes a multi-platform content generation system with calendar management and seo optimization, but lacks concrete procedural detail, failure handling, and actionable decision logic.

Generate skin swap prompts for model images, producing precise AI skin replacement instructions based on user-specified target skin tone/region while preserv...

VoxCPM2 — Tokenizer-Free TTS model guide. Covers installation, Python/CLI API (TTS/Voice Design/Controllable Cloning/Ultimate Cloning/Streaming), vLLM-Omni d...

OpenTUI — Zig-native terminal UI framework. Covers installation, renderer, components (Text/Box/Input/Select/Code/ScrollBox), Constructs declarative API, Fle...

Advanced AI-powered note assistant built into a note editor. Understands the three-layer context structure (Document/Block/Selection), distinguishes between"...

Use the Archiver library for streaming archive packaging in Node.js. Supports creating ZIP/TAR archives, appending content from streams, strings, buffers, fi...

Implement drag-and-drop interactions with @shopify/draggable. Supports Draggable (basic drag), Sortable (reordering), Droppable (drop zones), Swappable (swap...

markmap — Render Markdown as interactive SVG mindmaps. Use when users need to convert Markdown documents into mindmaps, generate HTML files via CLI, or navig...

Report only real risks, not manufactured panic. Covers injection, XSS, path traversal, insecure deserialization, authentication and authorization flaws, key...

LeaferJS — Canvas 2D Engine. Lightweight and high-performance Canvas 2D graphics engine, supporting multiple platforms such as Web, Worker, Node.js, and Mini...

Vercel CLI skill for deploying and managing Vercel projects from the terminal.

NestJS Node.js server-side framework development guide.

Helps agents conduct structured end-of-day review, reflection, and documentation. Provides capabilities to scan today's records, categorize activities, perfo...

Static site reproduction expert - Analyze target websites and manually code their structure, visual style, and basic interactions using pure HTML/CSS/JavaScr...

Prevents AI from giving up prematurely by exhausting all options, verifying fixes, troubleshooting proactively, and providing evidence before concluding.

SQLite database operations. Use this skill when users need to create, read, query, or modify SQLite databases (.db files).

Solves the memory fragmentation problem for Agents during session restarts, sub-agent boundaries, and cron/heartbeat isolation.

Facilitates simple multi-agent team collaboration with defined roles, task distribution, and versioned artifact management via a streamlined folder structure.

Automatically track tech company financial reports and generate investment summaries. Supports retrieving earnings calendars, market expectation comparisons,...

Create WeChat public account style business and technology articles with professional yet approachable tone.

For browser automation tasks, web data scraping, form filling, page screenshots, UI testing, and more.

Use mustache.js (logic-less Mustache templates) for any templating task in JavaScript/Node.js environments.

Aggregates and analyzes content trends across platforms to identify hot topics, user intent, content gaps, and generates data-driven article outlines.

Monitor WeChat window status, search contacts, and send messages on Windows via automated screenshots and input control.

Convert HTML DOM nodes to images (PNG/JPEG/SVG/Blob). Supports rendering options such as filter, backgroundColor, quality, and pixelRatio.

DOM capture engine that can export any DOM subtree as SVG, PNG, JPG, WebP, Canvas, or Blob, supporting inline styles, fonts, background images, pseudo-elemen...

Professional-grade Claude prompt optimization engine. Triggers when users submit raw prompts for optimization, refinement, or restructuring, or request "opti...

Generate various diagrams using Mermaid syntax (flowcharts, sequence diagrams, Gantt charts, class diagrams, state diagrams, pie charts, ER diagrams, mind ma...

Helps deeply understand consumer psychology and behavioral patterns based on psychological principles, providing strategic guidance on price anchoring, scarc...

Extract text from images using Tesseract.js (OCR). Supports multi-language recognition including Chinese and English, region recognition, character whitelist...

Based on a complete tarot knowledge base and systematic interpretation guide, provides accurate readings for various classic spreads along with comprehensive...

Node.js cron job scheduling with the `cron` npm package. Use when the user needs to schedule recurring tasks, create cron jobs, validate cron expressions, se...

Use WebTorrent to implement streaming BitTorrent client functionality in Node.js and the browser. Supports torrent downloading, seeding, magnet links, stream...

Fully automated collaborative code development pipeline for complex code development tasks. Must be used when users request code development, program writing...

Generate beautiful macaron-color cartoon illustration-style card images from text content. Supports various types such as book recommendation cards, concept...

AI Recommendation Engine Optimization (AEO/GEO) expert. Audit brand visibility on platforms such as ChatGPT, Claude, Gemini, and Perplexity. Analyze why comp...

Use Crawl4AI for web scraping and content extraction. Use when users need to scrape web content, extract structured data, convert web pages to Markdown, perf...

Send Web Push notifications from a Node.js backend using the web-push npm library (VAPID authentication, payload encryption). Covers generating VAPID keys, s...

Use markdown-it to render Markdown to HTML, configure plugins, custom rendering rules, syntax highlighting.

SEO automated content pipeline skill. Automates the entire workflow from competitor research and keyword mining to article generation and publishing.

Social media automation operations assistant. Automatically generate and publish content for Xiaohongshu, Weibo, and Twitter, with scheduled posting and inte...

Ink — React for interactive command-line apps. Build rich terminal UIs with React components.

Commander.js is the most popular command-line interface (CLI) framework for Node.js. Used to create professional CLI programs, supporting option parsing, sub...

Picomatch — A fast and accurate glob pattern matching library.

Create VS Code Copilot custom Agent (.agent.md) files.

Claude Agent SDK documentation — build production AI agents with Claude Code as a library in Python or TypeScript. Use when building, configuring, or debuggi...

Detailed Three.js 3D graphics reference covering scene setup, cameras, geometries, materials, lights, animations, controls, loaders, math utilities, and debu...

Clone, copy, rebuild, or reverse engineer any website. Use when users request to clone a website, copy a page, replicate a webpage, or recreate one from scra...

Personality Distiller — Automatically distill from name/vague requirement/link/existing Skill/local corpus into complete Agent persona file pack via deep res...

Automatically analyze and recommend optimal chart combinations based on data characteristics, generate beautiful interactive HTML dashboards (including line...

Match the most suitable type from 15 contrast gameplay styles (gender, age, identity, image, scene, object, role, consumption/economy, pre-post state, skill...

Bazi (Four Pillars Astrology) and Qi Men Dun Jia chart calculation and interpretation skills. Provides data analysis and cognitive science based calculations...