freestylefly/canghe-skills
30 skills14.3k installs12.7k starsGitHub
Install
npx skills add https://github.com/freestylefly/canghe-skillsSkills in this repo
1Manga Dramamanga-drama is a Python skill in freestylefly/canghe-skills that builds manga-style short dramas from a single protagonist image using the doubao-seedance-1-5-pro-251215 model. The manga_drama.py CLI analyzes the character, drafts scene scripts from five built-in templates (introduction, action, emotion, interaction, ending), and calls Seedance image-to-video at 1080p in 9:16 with 5-second scenes. Developers can run quick generate mode with --theme and --scenes or author JSON scripts via create-script and from-script, optionally pushing results to Feishu. ARK_API_KEY is required through .canghe-skills/.env or environment variables, and the skill depends on seedance-video-generation. A built-in manga style prompt preset enforces hand-drawn Japanese or Chinese comic aesthetics across every shot. Reach for manga-drama when you need repeatable vertical manga shorts for campaigns, character stories, or social clips instead of hand-editing every shot in a video editor or building a custom animation toolchain from scratch.759installs2Manga Style Videomanga-style-video is an agent skill from freestylefly/canghe-skills, listed on skills.sh with 510 installs, that converts storyboards or written scripts into consistent anime-style animated videos through AI agent workflows. The skill targets developers and content teams who need stylized motion content without manual frame-by-frame animation. Developers reach for manga-style-video when prototyping narrative demos, social clips, or explainer sequences from existing scripts or storyboard outlines. The workflow emphasizes visual consistency across scenes rather than one-off still image generation.618installs3Seedance Videoseedance-video is an agent skill for AI video generation and orchestration using Seedance APIs. Developers embed it in automated workflows to produce short-form video content, advertisements, and batch media assets without manually driving a separate video UI. The skill handles API integration patterns for agent-driven media pipelines where prompts and parameters flow from upstream planning steps. Reach for seedance-video when building content automation, marketing asset generation, or multi-step agent pipelines that output video files. The skill targets Seedance-specific API workflows rather than general FFmpeg or Remotion editing.564installs4Canghe Image Gencanghe-image-gen is an AI image generation skill from freestylefly/canghe-skills that runs `${SKILL_DIR}/scripts/main.ts` to call OpenAI, Google, DashScope (阿里通义万象), and Canghe providers for text-to-image and reference-image workflows with configurable aspect ratios. Sequential generation is the default, with parallel batches available on request, and project-level `.canghe` EXTEND.md preferences override defaults. Developers reach for canghe-image-gen when a coding agent must create, draw, or regenerate images inside a repository without switching to a separate design tool. Triggers include explicit image generation requests and reference-image edits during feature work.562installs5Volcengine Video Understandingvolcengine-video-understanding is a multimodal integration skill from freestylefly/canghe-skills, ranked 8 on Skills.sh with 445 installs. It equips coding agents to analyze video content through Volcengine's models—generating captions, summaries, and structured insights from footage. Developers invoke it when building features that ingest user uploads, surveillance clips, or marketing reels and need programmatic scene or content understanding without standing up a custom vision pipeline. The skill assumes Volcengine API access and fits agent workflows that combine video input with downstream code generation or data extraction.544installs6Canghe ComicCanghe Comic is an agent skill for indie creators and technical founders who explain complex ideas through visual serial formats—blog comics, launch threads illustrated as panels, or educational strips for newsletters. Before generating art prompts or scripts, the framework walks you through what readers must believe, which concepts deserve simplification, and which examples can be shown instead of narrated. It explicitly covers source bias, cultural period details, and prerequisite knowledge so you do not ship pretty pages that confuse newcomers. Solo builders use it to compress a whitepaper, podcast notes, or feature announcement into a paced story with identifiable character arcs and key beats. The output is analytical structure rather than finished PNGs; your design toolchain or illustrator workflow still runs afterward. It fits growth content calendars when you need differentiation beyond text posts, and it keeps scope bounded to analysis so agents do not drift into off-brand illustration without a plan.535installs7Canghe Xhs Imagescanghe-xhs-images is a Xiaohongshu infographic generator from freestylefly/canghe-skills that transforms long-form Markdown articles into 1–10 cartoon-style images tuned for Little Red Book engagement. It ships 10 visual styles and 8 layout templates, with CLI usage like `/canghe-xhs-images posts/article.md` and optional `--style notion` overrides for manual art direction. Developers and content engineers invoke it when users mention 小红书图片, XHS images, RedNote infographics, or 小红书种草 and need serialized social visuals instead of a single hero image. The skill auto-selects style and layout from content when flags are omitted, optimizing for Chinese platform scroll behavior.535installs8Canghe Cover Imagecanghe-cover-image is a Claude Code skill that instantly generates article cover images from markdown files or pasted content with five customization dimensions: type, palette, rendering, text, and mood. It combines 9 color palettes and 6 rendering styles, supporting cinematic 2.35:1, widescreen 16:9, and square 1:1 aspect ratios. Developers reach for canghe-cover-image when a blog post or doc needs a polished hero image without opening a design tool. CLI flags include --quick for skip-confirmation runs and explicit dimension selectors such as --type conceptual and --palette options.514installs9Canghe Markdown To Htmlcanghe-markdown-to-html is an agent skill from freestylefly/canghe-skills that converts Markdown files into styled HTML with inline CSS optimized for WeChat Official Account and other publishing platforms. The skill runs scripts/main.ts from the skill directory and supports syntax highlighting, math rendering, PlantUML diagrams, footnotes, alerts, and infographic blocks with responsive layout. Developers reach for canghe-markdown-to-html when a .md draft must become platform-ready HTML without hand-writing WeChat-compatible styles. Trigger phrases include markdown to html, convert md to html, and md转html, making it useful for bilingual technical blogging pipelines.509installs10Canghe Article Illustratorcanghe-article-illustrator is a skill from freestylefly/canghe-skills that generates cohesive illustrations and visual assets for long-form content. It helps engineering bloggers, newsletter authors, and documentation teams add scroll-stopping imagery while keeping color, style, and motif consistency across posts. Use canghe-article-illustrator when shipping technical articles or product changelogs that read better with matched header art, inline diagrams, or section dividers but you lack an in-house illustrator. The skill focuses on editorial visual systems—not app UI components or marketing ad creative suites.487installs11Canghe Infographiccanghe-infographic is a skill from freestylefly/canghe-skills that converts structured talking points, metrics, or feature lists into shareable infographic layouts. It targets developers and technical teams preparing launch assets, product explainers, or social posts who have the facts but not a design-ready composition. The workflow emphasizes layout from structured inputs—bullets, numbers, feature names—rather than freeform illustration. Reach for canghe-infographic when a release note, feature announcement, or metrics recap needs a visual summary for Twitter, LinkedIn, or blog embeds. It complements copywriting skills by producing visual artifacts suited to distribution, not replacing brand guideline systems or complex data-viz dashboards.483installs12Canghe Url To Markdowncanghe-url-to-markdown is a utility skill in the freestylefly/canghe-skills repository that launches Chrome via the DevTools Protocol, waits for JavaScript rendering, and writes structured Markdown files. The Bun CLI in scripts/main.ts supports auto capture on network idle and a --wait mode for login-gated or lazy-loaded pages, with a default 30000 ms timeout and optional -o output paths. html-to-markdown.ts runs five extraction strategies—Mozilla Readability, __NEXT_DATA__ parsing, JSON-LD, 14 CSS content selectors, and body fallback—then converts HTML with Turndown and GFM plugins. Output lands under url-to-markdown/<domain>/<slug>.md with six front-matter fields (url, title, description, author, published, captured_at). Developers use it to seed docs, archive articles, and feed summarization pipelines from SPAs and paywalled sessions they open manually.481installs13Canghe Format Markdowncanghe-format-markdown is a canghe-skills utility for normalizing, prettifying, and standardizing Markdown before commit or publish. It targets READMEs, documentation sites, changelogs, and agent-generated technical writing that drift into inconsistent heading levels, list formatting, or spacing. Developers reach for canghe-format-markdown when docs need a consistent house style without manual cleanup across many .md files. The skill fits pre-commit doc polish and post-generation cleanup for technical writing workflows.474installs14Canghe Slide Deckcanghe-slide-deck is a freestylefly/canghe-skills content skill that transforms markdown or pasted text into professional slide deck deliverables. The workflow builds a styled outline, generates per-slide image prompts, renders individual slide images, and merges outputs to PDF and PowerPoint via merge-to-pdf.ts and merge-to-pptx.ts scripts. CLI flags support --style presets, --audience targeting for executives or experts, --lang localization, and --slides counts with 8-25 recommended and a maximum of 30 slides. Partial modes include --outline-only, --prompts-only, --images-only, and --regenerate for specific slide numbers. Developers reach for canghe-slide-deck when validating product ideas, drafting pitch decks, or preparing stakeholder demo walkthroughs from existing outlines without manual slide design tooling.474installs15Remotion Best Practicesremotion-best-practices is a canghe-skills domain guide for programmatic video creation with Remotion and React. Instead of one long document, the skill indexes 28 focused rule files covering 3D with Three.js, animations, assets, audio, captions, charts, compositions, fonts, GIFs, Lottie, sequencing, timing, transitions, trimming, and video embedding with Mediabunny utilities. Developers reach for remotion-best-practices when building social clips, motion graphics, or data-driven renders and need Remotion-specific constraints such as frame-based useCurrentFrame animations instead of CSS transitions. The skill applies whenever Remotion code appears in a project, from scaffolding new compositions to optimizing export performance and caption workflows.469installs16Canghe Danger Gemini Webcanghe-danger-gemini-web is a canghe-skills integration that generates text and images through a reverse-engineered Gemini Web API when standard API access is insufficient. The skill ships scripts/main.ts as a Bun CLI entry point plus a TypeScript gemini-webapi client port, supporting text generation, image output, reference-image vision input, multi-turn sessions via --sessionId, and JSON responses with models such as gemini-2.5-pro. A required consent flow writes disclaimerVersion 1.0 acceptance to a platform-specific consent.json before first use. Developers reach for this skill when agent workflows need live page understanding, form interaction context, or image generation backends that other skills can call programmatically.468installs17Canghe Post To Wechatcanghe-post-to-wechat is an agent skill from freestylefly/canghe-skills that helps coding agents publish formatted articles and announcements to WeChat Official Accounts for Chinese-language distribution. The skill guides formatting, posting steps, and announcement structure so launch updates, feature releases, and promotional content reach subscribers on WeChat instead of staying as local markdown or English-only channels. Developers reach for canghe-post-to-wechat when a release needs WeChat Official Account coverage alongside other launch assets—particularly for products, APIs, or developer tools targeting mainland China users. It fits the distribution slice of a launch plan where the artifact is a live WeChat post rather than repository code. Pair it with general content or SEO skills for drafting, then invoke canghe-post-to-wechat when the remaining work is platform-specific publishing and audience reach on WeChat.467installs18Canghe Compress Imagecanghe-compress-image from freestylefly/canghe-skills is an agent skill focused on compressing and optimizing raster images before they ship to web pages, mobile applications, or automated content pipelines. The workflow targets smaller file sizes and improved load metrics while preserving perceptual quality for hero banners, thumbnails, icons, and editorial imagery. Developers invoke it when build artifacts or CMS uploads are too heavy, Lighthouse flags oversized images, or CDN bandwidth needs trimming ahead of a release. The skill fits into frontend and mobile delivery paths where manual batch tools are slow or inconsistently applied across a repo. Expect guidance and execution around lossy and lossless optimization choices, format selection, and batch processing of image directories rather than vector or video transcoding.464installs19Canghe Danger X To Markdowncanghe-danger-x-to-markdown is a content conversion skill from freestylefly/canghe-skills that transforms X (Twitter) posts and multi-post threads into clean, publication-ready Markdown. The skill strips platform formatting noise, preserves thread structure, and outputs Markdown suitable for static site generators, newsletter tools, documentation repos, and blog CMS imports. Developers reach for canghe-danger-x-to-markdown when repurposing launch threads, technical explainers, or changelog announcements into long-form blog posts or docs without hand-editing every line break and quote block. Output is a single Markdown file or thread-sectioned document ready for Hugo, Next.js MDX, Substack, or Obsidian vault ingestion.459installs20Canghe Post To Xcanghe-post-to-x is a Canghe agent skill for composing and publishing tweets or multi-post threads to X.com directly from an agent workflow. The skill enables social distribution, product announcements, and engagement updates without switching to a separate social client. Developers reach for canghe-post-to-x after shipping features or drafting release notes they want to broadcast on X. It automates the compose-and-publish step for X.com rather than building in-app UI or SEO pages. Use it when agent-driven workflows should post finished copy to X as part of a launch or growth routine.459installs21Obsidian Markdownobsidian-markdown is a documentation skill from freestylefly/canghe-skills that enables agents to create and edit valid Obsidian Flavored Markdown. The skill covers Obsidian-specific extensions on top of CommonMark, GitHub Flavored Markdown, and LaTeX math, including wikilinks, embeds, callouts, properties, tags, and YAML frontmatter. Developers reach for obsidian-markdown when editing .md files in Obsidian vaults, building team wikis with bidirectional links, or having agents generate research notes that must render correctly in Obsidian without syntax errors. Output is properly formatted .md files with valid wikilink syntax, callout blocks, image embeds, and frontmatter properties that open cleanly in Obsidian desktop and mobile.459installs22Find Skillsfind-skills is a Skill Development helper for Claude Code that discovers and installs community agent skills when a developer asks how to do X, whether a skill exists, or wants to extend agent capabilities. The skill routes queries like "find a skill for testing" or "can you do SEO" into a search-and-install workflow across the open agent skills ecosystem. Developers reach for find-skills when a task maps to an existing packaged workflow—documentation generation, SEO audits, test scaffolding, or domain-specific automation—rather than building from scratch. It pairs naturally with skill marketplaces and local SKILL.md directories, turning capability gaps into one-shot installs instead of manual repo hunting.456installs23Json Canvasjson-canvas is a canghe-skills agent skill for creating and editing valid JSON Canvas files with the .canvas extension used in Obsidian and other applications. Canvas files contain JSON with two top-level arrays for nodes and edges following the JSON Canvas Spec 1.0 from jsoncanvas.org. The skill supports text nodes, connections, groups, mind maps, and flowcharts when users mention Canvas files or need visual knowledge layouts inside Claude workflows. Developers reach for json-canvas when agent-generated diagrams must be saved as portable .canvas JSON rather than raster images or proprietary formats.455installs24Skill Creatorskill-creator in freestylefly/canghe-skills is an Anthropic-style guide for authoring effective agent skills. It explains skill anatomy: required SKILL.md with YAML name and description frontmatter, optional scripts for deterministic tasks, references for on-demand documentation, and assets for templates copied into outputs. Core principles emphasize concise context usage, matching degrees of freedom to task fragility, and progressive disclosure across three loading levels—metadata always, body on trigger, bundled resources as needed. The skill walks through understanding concrete examples, planning reusable contents, initializing directories, editing SKILL.md, and iterating from real usage. Reach for skill-creator when creating a new skill, converting a repeated prompt into a triggered workflow, or splitting an oversized SKILL.md into references so Cursor or Claude Code agents fire and follow domain procedures reliably.452installs25Obsidian Basesobsidian-bases in freestylefly/canghe-skills enables agents to create and edit Obsidian Bases—YAML-based .base files that define dynamic views of notes in an Obsidian vault. A Base file can contain multiple views, global filters, formulas, property configurations, and custom summaries including table views and card views. The skill activates when users work with .base files, mention Bases, or request filtered tables, rollups, and property-driven knowledge bases for ongoing PKM and research operations. Developers reach for obsidian-bases when vault note collections outgrow manual folders and need database-like querying without leaving Obsidian. Agents produce valid YAML configurations for views, filters, and formulas rather than ad hoc markdown tables. Because Bases read live vault properties, changes propagate automatically when note frontmatter updates—ideal for research pipelines tracking status, dates, and tags across hundreds of notes. Triggers include create Obsidian Base, table view with filters, card view rollups, and formula columns across vault properties.447installs26Release Skillsrelease-skills in freestylefly/canghe-skills provides a universal release workflow invoked via /release-skills that auto-detects project configuration without manual setup. Supported project types include Node.js (package.json), Python (pyproject.toml), Rust (Cargo.toml), Claude Plugin (marketplace.json), and generic projects with multi-language changelog support. Triggers include "release", "new version", "bump version", "push", and the Chinese equivalents 发布 and 推送. The skill handles version file detection, changelog updates, and distribution checks before sharing skills or packages with a team or registry. Developers reach for release-skills when repetitive release steps differ by language but follow the same bump-tag-publish rhythm. Agents skip guessing version paths because the workflow maps each project type to its canonical version file automatically. Common use cases include publishing canghe-skills entries, npm packages, Python libraries, Rust crates, and Claude marketplace plugins from one command without maintaining separate release scripts per repository.446installs27Douyin Downloaderdouyin-downloader is an agent skill in freestylefly/canghe-skills that downloads watermark-free Douyin (Chinese TikTok) videos and extracts speech-to-text transcripts from share URLs inside coding-agent workflows. Developers invoke it via /canghe-douyin-downloader with three subcommands: info to inspect a link, download to save video files with -o output paths, and extract to produce transcripts when DOUYIN_API_KEY is configured for SiliconFlow speech-to-text. The skill requires ffmpeg installed on the host and fits content research, archival, and downstream media-processing pipelines where agents must materialize Douyin assets as local files rather than manual browser downloads. Environment configuration centers on DOUYIN_API_KEY for transcript extraction while video download operates from share links alone.419installs28Flyworks Avatar Videoflyworks-avatar-video is a generative media skill for creating presenter-style avatar videos when products need human-feel explainer content without a video crew. It guides producing clips suited to product announcements, tutorial walkthroughs, onboarding sequences, and short social posts from scripts or talking points. Developers and technical content leads use it when documentation or UI copy exists but video assets lag behind release schedules. The skill emphasizes repeatable avatar-led production for engineering teams that ship features frequently and need consistent visual explainers tied to releases, demos, or in-app education flows.349installs29Paddleocr Doc Parsingpaddleocr-doc-parsing is an agent skill for developers who need to extract text and structure from scanned PDFs, images, and office documents using PaddleOCR. The skill focuses on turning visually-rendered documents into machine-usable artifacts such as text blocks, table content, and layout-aware chunks that can be indexed for search or fed into retrieval-augmented generation pipelines. Developers reach for paddleocr-doc-parsing when PDF text is not selectable, when invoices or forms must be parsed into structured fields, or when a RAG system needs high-quality OCR chunks with positional context. paddleocr-doc-parsing fits both one-off document parsing tasks and repeated ETL ingestion in production pipelines.343installs30Canghe TianyanchaTakes a company name and produces a single-file HTML dashboard covering profile, open roles, legal risk, financing, shareholders, IP, and investments. A user uses it for enterprise lookup and industry company analysis in the Kimi Work environment.129installs