alchaincyf/huashu-skills
21 skills44.1k installs26.4k starsGitHub
Install
npx skills add https://github.com/alchaincyf/huashu-skillsSkills in this repo
1Huashu Design设计哲学顾问 从20种风格中推荐3个方向并生成视觉Demo和AI提示词 当用户提到 设计风格 设计方向 配色方案 视觉风格 设计评审 推荐风格 时使用 name huashu-design description 设计哲学顾问 从20种风格中推荐3个方向并生成视觉Demo和AI提示词 当用户提到 设计风格 设计方向 配色方案 视觉风格 设计评审 推荐风格 时使用 Design Philosophy Consultant 像顶级设计企业那样理解需求 推荐设计哲学方向 生成AI提示词 并进行专家级评审 Core Principles 1 约束哲学而非形式 定义 为什么这样设计 而非 长什么样 2 深度理解优先 先理解用户真正要传达什么 再推荐风格 3 设计是概率性的 好的约束产生多样化的高质量结果 4 评审如乔布斯 高标准 但给出可执行的改进路径 When This Skill Triggers 设计前 Design Direction 任何涉及视觉呈现的输出需求 网页 落地页 官网 App界面 原型 PPT Keynote PDF报告 白皮书 信息图 海报 配图 封面 设计后 Design Critique 当视觉设计产出完成后自动启动评审 Workflow 设计前 Phase 1-6 Phase 1 深度理解需求 提问 目标受众 核心信息 情感基调 输出格式 一次最多3个问题 需求已清晰则跳过 Phase 2 顾问式重述 100-200字 用自己的话重述本质需求 受众 场景 情感基调 以 基于这个理解 我为你准备了3个设计方向 结尾 Phase 3 推荐3套设计哲学 每个方向包含 要素 说明 风格名称 必须含设计师 机构名 如 Kenya Hara式东方极简 不能只说 极简主义 为什么适合 50-100字 连接需求与该设计师的哲学 核心特征 3-4条标志性视觉要素 气质关键词 3-5个 代表作参考 可选 方便用户搜索理解 差异化策略 必须遵守 3个方向必须来自 3个不同流派 形成明显的视觉反差 流派 视觉气质 适合作为 信息建筑派 01-04 理性 数据驱动 克制 安全 专业选择 运动诗学派 05-08 动感 沉浸 技术美学 大胆 前卫选择 极简主义派 09-12 秩序 留白 精致 安全 高端选择 实验先锋派 13-16 先锋 生成艺术 视觉冲击 大胆 创新选择 东方哲学派 17-20 温润 诗意 思辨 差异化 独特选择 推荐组合示例 Pentagram编辑风 信息建筑31.1kinstalls2Huashu Md To Pdfhuashu-md-to-pdf is a huashu-skills generator that runs convert.py to turn Markdown sources into clean, professionally styled PDFs. The pipeline reads a file, extracts YAML metadata, builds a table of contents—example runs report 5 main chapters and 12 sub-chapters—renders HTML, and writes a PDF with optional custom output names, titles, and authors. Developers reach for it when README files, tech guides, or whitepapers must ship as PDF attachments without manual Pandoc or LaTeX setup. Commands support basic conversion, custom output filenames like 产品说明书.pdf, and full metadata overrides from the CLI.940installs3Huashu Douyin Scripthuashu-douyin-script is a six-step Douyin script workflow from alchaincyf/huashu-skills that downloads 1–5 competitor videos with yt-dlp, analyzes them via the Gemini API, extracts viral formulas, and generates scripts plus shot breakdowns with AI tone review. Required tooling includes uv, yt-dlp, a GEMINI_API_KEY, and a Chrome session logged into Douyin for download cookies. Developers trigger it for 抖音脚本, 爆款拆解, 竞品分析, 带货脚本, 千川素材, and 种草脚本 requests. Reach for this skill when you need a repeatable pipeline from competitor URLs to structured Chinese short-video scripts instead of manual imitation.849installs4Huashu Proofreadinghuashu-proofreading is a systematic Chinese proofreading skill for drafts that read too machine-generated. It executes three sequential passes—content accuracy (facts, logic, structure), style de-AIing across six AI-voice pattern categories, and final polish—aiming to keep AI detection below 30% while preserving a practical, warm author voice. Triggers include phrases like 审校, 降低AI味, or 太AI了 after an initial draft is complete. The first pass verifies data, product names, and claims via search; the second removes stock phrases and robotic cadence; the third tightens readability. Developers writing Chinese blog posts, newsletters, or docs reach for this skill when automated detectors flag high AI scores or copy lacks experiential tone.757installs5Huashu Slideshuashu-slides is an alchaincyf/huashu-skills skill that instantly transforms markdown documents, technical specs, or prior agent output into polished slide decks suitable for presentations. Developers reach for huashu-slides when a written artifact already exists and needs deck formatting without manual copy-paste into slide tools. The skill targets engineers presenting architecture reviews, sprint demos, or stakeholder updates where source content lives in markdown or agent-generated text. It reduces friction between documentation and presentation deliverables. Limited public readme text is available beyond the stated markdown-to-slides conversion scope.722installs6Huashu Wechat Imagehuashu-wechat-image is a WeChat article image workflow skill with a mandatory propose-then-generate flow: understand content, present two to three design directions, user selects, then render and preview. It supports AI generation for creative visuals and HTML rendering for text-precise graphics. Documented sizes include 1800×766 px headline covers (2.35:1), 1920×1080 px body images (16:9), and additional aspect ratios with matching Playwright viewport flags such as --viewport-size=1800,766. Developers reach for this skill when building 公众号配图, covers, or infographics that must meet WeChat layout constraints before upload.646installs7Huashu Data Prohuashu-data-pro is an alchaincyf/huashu-skills workflow that covers end-to-end data processing, insight generation, report writing, PPT creation, and visualization from Excel and ad campaign data. The skill defaults to interactive HTML reports with ECharts charts and PDF export, switches to slide-style HTML for PPT requests, and enforces honest data handling with zero baselines and cited sources. Developers and analysts reach for huashu-data-pro when triggers mention 分析数据, Excel, ROI, 周报, or 可视化 and they need structured deliverables instead of ad hoc chart snippets. Use it for campaign postmortems, formula generation, and executive-ready reporting from messy tables.643installs8Huashu Researchhuashu-research is a structured web research agent skill that persists findings to disk before searching begins. Step 1 creates `_knowledge_base/research-<主题>-<YYYYMMDD>.md` with goals, key questions, and empty findings sections. Step 2 appends each WebSearch result with source URLs immediately. Step 3 writes a stage summary after every 3 searches. Step 4 produces a final briefing. The skill triggers on research, search, or information-gathering requests and follows source-priority rules from SHARED-RULES.md. Developers reach for huashu-research when multi-round web investigation must survive long agent sessions, such as competitive analysis, technology evaluations, or pre-writing research for technical articles.636installs9Huashu Xhs Imagehuashu-xhs-image is a Xiaohongshu (Little Red Book) image workflow skill that defaults to Gemini AI generation for note covers and carousels, using HTML screenshots only as a fallback for precise data tables. The skill enforces a proposal-first pipeline: understand content, present 2-3 design directions, wait for user selection, then generate and preview before upload—never skip straight to images. huashu-xhs-image encodes a design taste profile favoring warm paper textures, handwritten fonts, hero-sized keywords, and mobile vertical layouts while rejecting cyber-neon palettes, excessive whitespace, and flat HTML-slide aesthetics. Developers and content engineers reach for huashu-xhs-image when automating 小红书配图, 封面, or carousel assets with agent-guided brand consistency rather than one-shot generic AI art.636installs10Huashu Script Polishhuashu-script-polish is a video script oral-editing agent skill that removes written-language tone so scripts sound natural on camera. It runs a three-pass review: content accuracy and demo feasibility, oralization with short sentences under 20 characters and conversational fillers, then demo cut annotations for screen switches and close-ups. The skill targets 150-180 characters per minute pacing and replaces phrases like "接下来我们将进行" with direct speech like "我们来". Developers reach for huashu-script-polish after drafting a technical tutorial or product demo script that reads well on paper but feels stiff when spoken aloud. It triggers on requests for oralization, script review, or de-formalization.635installs11Huashu Topic Genhuashu-topic-gen is a huashu-skills step that instantly produces 3-4 distinct WeChat 公众号 article topic directions, each with a title, core angle, outline, workload star rating, and balanced pros/cons analysis. It activates on 选题, 写什么, 文章方向, or 题目建议 and runs after a brief is saved in the huashu article workflow (Step 3). Four angle types—深度评测型, 实战教程型, 洞察观点型, and 案例拆解型—include workload guidance from one to three stars. The skill waits for explicit user selection instead of assuming a pick. Developers and technical writers reach for huashu-topic-gen when planning Chinese-language WeChat posts and need structured topic comparison before drafting.630installs12Huashu Speech Coachhuashu-speech-coach from alchaincyf/huashu-skills coaches presentation preparation based on Patrick Winston's MIT How to Speak methodology, referenced from a YouTube lecture with over 15 million views. The skill helps developers structure offline training workshops, optimize Bilibili tutorial scripts, review meetup talk drafts, and audit PPT or Keynote slides for information density and common slide mistakes. Winston's formula treats speech quality as a function of knowledge, practice, and minimal innate talent. Developers reach for huashu-speech-coach when preparing openings, closings, interactive segments, pacing hooks, or reviewing presentation outlines in Chinese or English technical speaking contexts.616installs13Huashu Video Outlinehuashu-video-outline is an alchaincyf/huashu-skills agent skill that quickly produces two or three distinct video script outline options, each with an attention-grabbing title, cover thumbnail suggestion, core angle, timed section breakdown, and estimated duration between five and eight minutes. Outlines follow a five-part structure—opening hook, two core modules, surprise beat, and closing call-to-action—with demo checkpoints and explicit advantage versus disadvantage notes per option. Title strategy applies five contrast formulas covering quantity, price, result, time, and capability comparisons to strengthen click-through potential. Developers reach for huashu-video-outline after saving a video brief or when users request outline options during Step 3 of a video creation workflow inside Claude Code or Cursor. Output suits developer advocates recording framework tutorials, release walkthroughs, or tool comparisons who need structured scripts before opening a camera or screen recorder.605installs14Huashu Article Edithuashu-article-edit is a standardized article editing skill in alchaincyf/huashu-skills that prevents long editing sessions from losing work when agent context truncates. The workflow reads the full target article, lists every modification with location and type, waits for user confirmation before editing, then applies incremental changes with immediate file saves after each batch. Progress reports appear every 3-5 completed items showing done and pending modifications. Developers reach for huashu-article-edit when editing, restructuring, or tone-adjusting existing articles or documentation that require many scoped changes across a single long document.604installs15Huashu Info Searchhuashu-info-search is an alchaincyf/huashu-skills agent workflow that automatically searches official docs, tech media, and community forums when developers ask about new products or technologies—especially releases within the last three months. It cross-verifies claims across channels, prioritizes authoritative sources such as TechCrunch, The Verge, MIT Technology Review, Reddit, and Hacker News while ignoring stale Zhihu or Baidu posts from before 2025 and content older than six months. Verified findings save to a _knowledge_base/ directory with source URLs and timestamps for later agent retrieval. Search strategies combine topic-plus-year keywords, competitor comparisons, and concept explanations from official announcements, pricing pages, and community threads. Developers reach for huashu-info-search when agents must answer what shipped recently, compare rival frameworks, or compile citable research briefs without manual tab switching. Triggers include phrases about latest information, new products, or searching reference materials in Chinese or English prompts.602installs16Huashu Article To Xhuashu-article-to-x is a Marketing & SEO skill from alchaincyf/huashu-skills that repurposes finished long-form articles into short social posts for X (Twitter), Weibo, and Xiaohongshu. The skill targets 3000–5000-character source articles and outputs 200–500-character posts that preserve core arguments, conversational tone, and engagement hooks. Three entry angles—punchy thesis openers, data or results leads, and direct value propositions—guide structure when users ask to shorten articles or generate platform promotion. Developers and technical writers reach for huashu-article-to-x after publishing a blog or essay and needing platform-native teasers without rewriting from scratch.599installs17Huashu Image Uploadhuashu-image-upload is a content-illustration skill for Markdown articles that analyzes draft text, selects 5-8 images (minimum 3, maximum 10), sources them from public-domain archives, free stock libraries, or the doubao-seedream API, uploads each file to ImgBB for a permanent URL, and inserts `` blocks at cover and section breakpoints. Developers reach for huashu-image-upload when an article is structurally complete but lacks visuals, or when triggers like 配图, 插图, or 上传图片 appear in the request. The skill validates every hosted link before returning the updated Markdown. It targets editorial workflows rather than app UI asset pipelines.592installs18Huashu Material Searchhuashu-material-search is a content-research skill that queries a bundled personal archive of 1800+ Jike posts (May–October 2025) to inject authentic anecdotes, product opinions, and case material into drafts. The archive spans themed CSV sources—including 504 AI-programming-tool posts, 200 product-development notes, 150 AI-model reviews, and 100 growth posts—and returns keyword matches plus rewrite guidance for expanding short posts into long-form logic. Developers reach for huashu-material-search when copy reads generic or needs verifiable personal examples for essays, newsletters, scripts, or social threads. It complements generic LLM drafting by anchoring output in curated, author-specific source text rather than inventing experiences.587installs19Huashu Agent Swarmhuashu-agent-swarm is a Claude Code skill inspired by Nicholas Carlini's 16-instance C compiler experiment. It runs multiple independent Claude CLI agents in tmux sessions that claim tasks from TASKS.md, write code, and push through git with no master orchestrator. Setup runs bash scripts/setup_project.sh to generate AGENT_PROMPT.md, TASKS.md, and a current_tasks/ claim directory inside an existing git repository. Developers reach for huashu-agent-swarm when a monolithic agent session cannot keep pace with a large refactor, greenfield build, or parallel feature track. Default configuration starts 8 agents, and prerequisites include tmux, the Claude CLI, and a git repo with defined test commands.584installs20Huashu Video Checkhuashu-video-check is a video optimization skill from alchaincyf/huashu-skills that reviews titles, cover images, and content handoff using MrBeast-style strategies to raise click-through rate and watch time. It checks three dimensions: title strong-contrast formulas (quantity, price, time, and similar patterns), thumbnail result-display priority, and opening-hook structure spanning confirmation, mid-video surprise, and payoff delivery. Developers trigger the skill when scripts are drafted, CTR is low, or sessions mention titles, covers, thumbnails, or watch duration in Chinese or English contexts. huashu-video-check outputs standardized, reusable封标承接 guidance instead of generic creative feedback.584installs21Huashu Prompt Savehuashu-prompt-save is an alchaincyf/huashu-skills agent skill that automatically classifies pasted prompts into five categories—technical development, content creation, instructional design, product consulting, and general tools—then saves them as numbered markdown files and updates a central INDEX.md. Trigger phrases include save prompt, record prompt, or organize prompt in Chinese or English contexts. The skill extracts purpose, source, applicable scenarios, and key characteristics before writing files named prompt-[序号]-[标题].md into the matching category folder. Developers reach for huashu-prompt-save when accumulating battle-tested prompts from chat sessions, network sources, or custom edits and need a searchable local library instead of scattered notes. It handles classification rules via keyword tables, tag assignment, and index maintenance so repeated prompts stay discoverable across coding, writing, teaching, and product workflows.576installs