Most tasks in the self-media production line can now be handled by Skills. From topic selection and layout to video editing and publishing, ready-made solutions exist for every step.
Terminal Agents like Codex, Claude Code, and Openclaw have seen a surge of new Skills this year. The principle is simple: write a fixed workflow into a file, install it, and the Agent works according to your rules.
Below is a list organized by workflow stage, with repository links and brief descriptions. GitHub star counts are as of October 8, 2026, for reference only—high stars don't guarantee fit; choose what works for you. Some low-star tools are included because I use them personally and find them effective.
Each entry ends with a copy-paste command to install the Skill directly into your agent.
Table of Contents
- 1. Topic Selection Brain
- 2. Article Layout & Images
- 3. Video Skills
- 4. Xiaohongshu (RED) Operations
- 5. WeChat Official Account Operations
- 6. Douyin (TikTok China) Operations
- 7. Bilibili & YouTube Operations
- 8. Twitter/X Operations
- 9. Optional Video Skills
- Quick Reference Table
- Final Thoughts
1. Topic Selection Brain
① Global Strategist · self-media-content-workflow ⭐521
Oversees account positioning, topic planning, content production, and data review—the commander-in-chief of the entire production line.
A modular self-media content production suite that takes vague requirements through to delivery and archiving: creative briefs, account strategy, trend and competitor research, platform-native copy, short video plans, digital human production, multi-platform publishing, and data review.
Three key features: It doesn't bind to specific models or browsers, discovering available capabilities at runtime. It forces stops for confirmation at five critical points (direction, platform, title, final draft, publish authorization), defaulting to drafts only and never auto-publishing. Key numbers must have sources; no fabricated data or earnings. If you want a master control skill, start here.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/yanhua1010/self-media-content-workflow
``
② Trend Radar · creator-buddy ⭐1.6k
Scans major platforms 24/7 for rising trends and real pain points in comments, avoiding echo chambers.
The slogan is "From one topic to a publishable finished product." It finds cross-platform content, analyzes creators, and tracks viral trends, turning "what to write today" from guesswork into evidence-based decisions. Works with Claude Code, Codex, Cursor, Openclaw, and Hermes.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/SpaceZephyr/creator-buddy
``
③ Benchmark Crusher · blogger-distiller ⭐694
Frame-by-frame breakdown of top competitors' topic strategies, content structures, and monetization tactics.
Input a blogger's account, and it distills their proven methods from real posts into your AI, creating an on-demand content coach. It can also distill your own work—lay out your last thirty posts, and you'll see what you repeat and what you lack. Switch benchmark bloggers without starting over.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/otter1101/blogger-distiller
``
2. Article Layout & Images
④ Layout Money Printer · baoyu-skills ⭐26.4k
Official skills from Baoyu! Auto-generate beautiful images, SVG covers, and one-click multi-platform formatting.
Baoyu's shared collection of twenty-plus skills. Common ones for self-media: baoyu-cover-image for cover art, baoyu-article-illustrator for body illustrations, and baoyu-post-to-wechat for WeChat publishing.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/JimLiu/baoyu-skills
``
Official CLI: npx skills add jimliu/baoyu-skills
Warning: Don't install all of them. The author notes in the README that installing all twenty-plus skills bloats the context, slowing down the Agent. Install only what you need.
⑤ Card Generator · guizang-social-card-skill ⭐7.4k
By Guizang! Turn long articles into high-aesthetic Xiaohongshu image sets and WeChat covers with one click.
Generates Xiaohongshu image-text sets, Live Photo dynamic cards, and WeChat cover pairs (21:9 + 1:1) from articles, copy, screenshots, subtitles, or photos. Includes two visual systems: Electronic Magazine style for narrative/travel/reading, and Swiss International Style for reviews/data/tutorials. 28 layouts, 10 themes, single-file HTML to PNG conversion.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/op7418/guizang-social-card-skill
``
Official CLI: npx skills add https://github.com/op7418/guizang-social-card-skill --skill guizang-social-card-skill
⑥ Article Illustrations & Covers · Punk-Skill ⭐1.2k
AdrianPunk's visual generation skill set.
Generate cover images from articles, or create avatar images from photos of people, pets, or objects.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/adrianpunk/Punk-Skill
``
Check License First: Source code is public since v2.0.0, free for personal non-commercial use. Commercial use (enterprises, agencies, client projects, paid services, marketing, ads, brand propagation, product integration, commercial workflows) requires prior written authorization and payment. Fine for hobbyists, not for freelancers selling services.
⑦ Body Illustrations · ian-xiaohei-illustrations ⭐12.4k
Transform judgments, processes, states, and metaphors in Chinese articles into white-background, hand-drawn, quirky but clean body illustrations.
Xiaohei's style: 16:9 landscape, pure white background, hand-drawn, sparse red/orange/blue Chinese annotations. It first understands cognitive anchors in the article, then illustrates one judgment or process—not just pasting a mascot in the corner. Suitable for knowledge content, methodologies, and AI workflows. Not ideal if you want commercial illustration or refined flat design styles.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/helloianneo/ian-xiaohei-illustrations
``
⑧ Write Then Publish · write-then-publish ⭐561
Once written, no need to reformat everywhere.
One body text simultaneously generates Xiaohongshu image-text cards and WeChat long-form articles. Switching modes changes output format, not rewriting. Images upload and arrange instantly, with cropping and alignment adjustments. Two-way sync with Obsidian: writes back Markdown and attachments if directory permissions allow, otherwise degrades to ZIP for manual extraction. It doesn't write content for you; it solves the "last mile" after writing.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/fxyadela/write-then-publish
``
3. Video Skills
⑨ Long Video Scalpel · video-to-subtitle-summary-skill ⭐213
Instantly extract subtitles and key points from YouTube/Bilibili/Podcasts, the core extractor for converting long videos to text/images.
Input links from Douyin, Xiaohongshu, Bilibili, YouTube, or local audio/video files to get subtitles plus AI summaries. Defaults to local faster-whisper transcription (free, no uploads). YouTube uses yt-dlp to grab manual or auto subtitles directly, skipping ASR. Cloud transcription via Volcano Engine is optional but requires configuring tokens.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/imlewc/video-to-subtitle-summary-skill
``
⑩ AI Video Storyboard Artist · seedance-prompt-skill ⭐2.9k
Input one sentence, automatically output video scripts with camera movements, detailed storyboards, and prompts.
Convert ideas into structured Chinese prompts ready to paste into Jimeng Seedance 2.0. Among its ten capabilities, consistency control (locking characters/products with reference images), camera movement replication (copying shots/rhythms from reference videos), and effect template replication are where it beats manual prompting.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/songguoxs/seedance-prompt-skill
``
Note: Original draft said seedance-skills, but the verified repo is seedance-prompt-skill. Check if they are the same before installing.
⑪ Talking Head Motion Studio · video-talkcraft ⭐1.4k
Agent skill for talking head videos: word-level voice sync · 108 motion recipe cards · seven-layer anti-PPT shot system · triple verification.
Give it a script and a finished voiceover. It aligns timestamps to the word level locally, writes each semantic beat into SHOTBOOK storyboards, and renders the final video using Remotion—dynamic text cards, evidence screenshots, camera moves, plain subtitles, cinematic sound effects, all locked to the voice track. The sibling project video-shotcraft handles other video types; this one specializes in talking heads.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/Vincentwei1021/video-talkcraft
``
License: PolyForm Noncommercial. Another Pitfall: One node_modules folder is ~760 MB, including a 190 MB headless browser. The author originally installed separate copies per video, consuming 3.9 GB for six projects. Now switched to symlink sharing—run link-runtime.sh before connecting old projects.
⑫ Multi-Platform Video Publisher · video-publisher-skill ⭐229
Xiaohongshu · Douyin · Bilibili · WeChat Channels · YouTube, parallel upload, verify upon completion.
Prepares videos, titles, and covers into verified multi-platform drafts, resuming interrupted uploads. Stops before the final publish button by default—only acts when you explicitly say "publish to which platform," then re-verifies the draft. Draft preparation, uploading, and long-term account configuration do not count as authorization. Among similar tools, it is the strictest about "don't publish randomly."
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/oil-oil/video-publisher-skill
``
4. Xiaohongshu (RED) Operations
⑬ Viral Content Monitor · content-boom-monitor ⭐106
Turns "search keywords for virals" and "watch benchmark accounts for anomalies" into stable data flows.
Monitors keywords on both Xiaohongshu and Douyin, and watches recent posts from specified benchmark accounts. Deduplicates across sources, calculates platform-differentiated engagement values, computes relative performance R against the account's historical median, and labels results as T1 Potential / T2 Viral / T3 Phenomenal. Results settle into Feishu multidimensional tables; humans only handle the worth-looking-at entries.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/xintu1314/content-boom-monitor
``
Prerequisite: Uses Just One API for public data collection; requires registration and key acquisition, not plug-and-play.
⑭ Xiaohongshu Operator · xiaohongshu-ops-skill ⭐2.4k
Full coverage from viral copy generation and SEO keyword embedding to publishing interactions and account diagnosis.
Works with Openclaw to independently operate a Xiaohongshu account: recommendation feed analysis, account diagnosis, topic inspiration, viral replication, auto-publishing, comment replies. Based on browser automation CDP; scan QR code once for login, no repeated verification afterward. Viral replication takes a viral post link, analyzes why it went viral, and generates a similar image-text post.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/Xiangyu-CAS/xiaohongshu-ops-skill
``
5. WeChat Official Account Operations
⑮ Full Pipeline · wewrite ⭐3.4k
From trend scraping to WeChat draft box, run the entire content pipeline with one sentence.
Say "Write a WeChat article," and it scrapes trends, evaluates topics, searches real materials, drafts in your persona/style, and runs an editorial review. After the body text, you can append anytime: generate covers/illustrations, preview 18 layout themes, push to draft box, rewrite for other platforms. Original text is preserved; illustrations are generated as independent copies. Manually edit, then tell it to "learn my edits," making the next article more like you.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/imraywang/wewrite
``
⑯ Article Distributor · Wechatsync ⭐6.4k
One-click sync WeChat articles to Zhihu, Toutiao, Juejin, Xiaohongshu, CSDN, and 29+ other platforms.
Chrome extension, not a crawler, no simulated login, no intermediate server. Uses your browser's existing login state to call official Web APIs, identical to manual web publishing; data never leaves your device. Defaults to syncing as drafts; you confirm before publishing.
The extension is for browsers. To let Agents call it directly, official paths include Skill and MCP. Skill is recommended—install it, then say "Sync this article to Juejin and Zhihu" in Claude Code, and it publishes.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/wechatsync/Wechatsync
``
Official CLI (Claude Code):
``bash
/plugin marketplace add wechatsync
/plugin install wechatsync
``
Prerequisites: Skill is just the entry point; the extension must be installed first—command line communicates via localhost. Also generate a token in extension settings and fill it into environment variables. All three steps are required.
6. Douyin (TikTok China) Operations
⑰ JianYing Automation · jianying-headless ⭐3.0k
Local automation for JianYing Pro on macOS.
Write a structured editing plan, and it generates editable JianYing drafts, modifies multi-track projects in isolated copies, and calls the local JianYing engine to export MP4. Suitable for batch draft creation, Agent-assisted editing, and project handoffs.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/mcncarl/jianying-headless
``
Two Prerequisites: Runs only on macOS. Primarily adapted for JianYing 11.5.0, compatible with 11.4.2. Not an official SDK; version mismatches cause issues.
⑱ Yichen's JianYing Editing · yichen-skills ⭐4.3k
A creator skill set: Writing, X content slicing, X Articles draft publishing, WeChat material scraping.
Inside, yichen-jianying-edit is the JianYing editing chain. The writing and X-related skills in the full set will be used again in the "Twitter/X Operations" section below.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/mcncarl/yichen-skills
``
License Reminder: Personal learning and non-commercial use only. Client projects, paid products, internal company deployment, packaged listings, course bundles require prior written authorization.
Original Tweet: <https://x.com/bbkirstry/status/2104030018544566539>
7. Bilibili & YouTube Operations
Bilibili and YouTube are long-video platforms; dedicated skills are scarce. Usable options are mostly general video tools plus platform publishing. Both platforms are discussed together.
Subtitles & Topics · video-to-subtitle-summary-skill ⭐213
See Item 9 in "Video Skills." Extracting subtitles and summaries from Bilibili videos is the first step for reposting and topic finding. YouTube links go directly through yt-dlp, prioritizing manual subtitles, falling back to auto subtitles, without downloading video or running ASR—the most efficient path.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/imlewc/video-to-subtitle-summary-skill
``
Upload Distribution · video-publisher-skill ⭐229
See Item 12 in "Video Skills." Both platforms are in its verified list. Bilibili follows the generic flow like Xiaohongshu/Douyin. YouTube is the only one with complete single-platform verification—metadata, 16:9 thumbnails, private visibility, post-publish checks are all verified. Prioritize it for overseas publishing.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/oil-oil/video-publisher-skill
``
Full Chain Hosting · self-media-content-workflow ⭐521
See Item 1 in "Topic Selection Brain." Multi-platform video publishing covers Douyin, WeChat Channels, Xiaohongshu, and can hook up Bilibili.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/yanhua1010/self-media-content-workflow
``
Overseas Content Line · yichen-skills ⭐4.3k
The author himself focuses on going overseas; YouTube is his primary publishing channel. The writing and slicing chains in this skill set connect directly to YouTube content.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/mcncarl/yichen-skills
``
8. Twitter/X Operations
⑲ Publish X Articles · x-article-publisher ⭐5
Super useful X long-form publishing skill, perfectly supports markdown file pushing, tables, code blocks, images!
Feed Markdown into X Articles drafts, fully automated headless import. Playwright plus system Chrome, no API keys needed. Only fills drafts, never auto-publishes—you manually click Publish.
This is my own maintained project, originating from xPoster, forked through punk2898's v4.1.0, with changes recorded in NOTICE and LOCAL_CHANGES.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/xaiwind/x-article-publisher
``
Official CLI (Claude Code):
``bash
git clone https://github.com/xaiwind/x-article-publisher.git ~/.claude/skills/x-article-publisher
cd ~/.claude/skills/x-article-publisher && npm install
``
For Codex / DeepSeek Harness, change the path to ~/.agents/skills/. Do not use `--headless` on first run; the browser opens for X login, and the session persists. After installation, say "Help me publish the article to Twitter" to invoke it.
For a more mature general solution, check wshuyi/x-article-publisher-skill ⭐869, where v1.2.0 supports dividers, table-to-image conversion, and Mermaid.
⑳ X Slicing & Drafts · yichen-skills ⭐4.3k
See Item 18 in "Douyin Operations." Cutting long content into multiple X posts and publishing X Articles drafts are both in this set.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/mcncarl/yichen-skills
``
9. Optional Video Skills
Previous items targeted specific stages. This section differs: these are general video production tools, platform-agnostic, take as needed.
HyperFrames · heygen-com/hyperframes ⭐58.9k
Write HTML, render video.
Open-source video rendering framework that renders HTML, CSS, media files, and addressable animations into deterministic MP4 files. Local CLI available, also installable as a Skill for Agents. Repo includes 21 Skills loaded on demand; /hyperframes is the entry router—you say "Make a 10-second product intro," and it picks the workflow. Apache 2.0 license.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/heygen-com/hyperframes
``
Official CLI (Claude Code):
``bash
claude plugin marketplace add heygen-com/hyperframes
claude plugin install hyperframes@hyperframes
``
Pitfall: Don't install all 21 at once. Author says the core group is enough; other workflows are pulled by the router on demand. Run npx hyperframes skills update to install the core set, cleaner than adding --all.
Remotion Core · remotion-dev/remotion ⭐62.5k
The item below is the rule; this one is the core.
Remotion is a framework for writing videos with React. The Skill above only manages rules; without installing the core, rules cannot be implemented.
Installation method differs—it's not a Skill installed outside the project, but a package installed inside your project:
``bash
npx create-video@latest
``
For existing projects:
``bash
npm i remotion @remotion/cli
``
License is custom, not standard open source. Free for individuals, non-profits, and for-profit companies with ≤3 people, including commercial use. Larger companies must buy licenses.
The official repo itself includes 60 Skills, but those are for maintaining Remotion—adding effects, releases, bug fixes, changing website demo videos. If you're just making videos, you likely don't need to install them.
Remotion Video Creation · remotion-video-creation ⭐275k (Collection Total)
Write videos with React, condensing best practices into 29 rules.
Remotion is a framework for writing videos with React. This Skill condenses its best practices into 29 rules covering 3D, animation, audio, subtitles, charts, transitions. After installing, when you say "Make an animation with Remotion," the Agent won't write nonsense.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/affaan-m/everything-claude-code
``
Note: It's in the affaan-m/everything-claude-code collection, totaling 292 Skills, 275k stars. The command above installs all 292. If you only want this one, clearly tell the Agent to install only the skills/remotion-video-creation/ directory.
Manim Animation · manim-video ⭐275k (Collection Total)
For explanatory animations, draw them with code.
Manim is the animation engine behind 3Blue1Brown. This Skill lets it write Manim scenes for you: technical concepts, function curves, system architecture diagrams, product walkthroughs. Suitable for explanatory videos, not for talking heads or Vlogs.
Copy-paste to install:
``bash
帮我安装这个仓库的全部skill：https://github.com/affaan-m/everything-claude-code
``
Same repo as ㉒, same installation caution: don't install everything if you only want this.
video-use · browser-use/video-use ⭐28.4k
Drop materials into a folder, chat with the Agent to edit.
MIT license. Does practical things: removes filler words (umm, uh, false starts) and gaps between cuts, auto-color grades, adds 30ms audio fade-in/out at each cut point (avoiding pops), burns subtitles in your style (default two-word groups capitalized). Animation overlays are delegated to HyperFrames, Remotion, Manim, or PIL, spawning parallel sub-Agents, one per animation. After rendering, each cut point self-checks before showing you.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/browser-use/video-use
``
JianYing Editor · jianying-editor-skill ⭐3.8k
Let the Agent directly modify JianYing drafts, handling copy, voiceover, subtitles, effects to export in one go.
MIT license. Targets domestic JianYing Pro, supporting draft generation on Windows and macOS. Automatic export is stable only on Windows, relying on Windows UI Automation, most reliable with JianYing 5.9 or lower. macOS can only generate drafts; final export must be done manually in JianYing. CapCut international and mobile versions are not supported.
Copy-paste to install:
``bash
帮我安装这个skill：https://github.com/luoluoluo22/jianying-editor-skill
``
Difference from ⑰: ⑰ is macOS native, calling the local JianYing engine to export finished videos. This one is reversed: strong on Windows, macOS only outputs drafts. If you're all Mac, think carefully about which segment you need first.
Quick Reference Table
Stage | Skill | Stars | One-liner |
|---|---|---|---|
Master Control | self-media-content-workflow | 521 | From brief to delivery archive |
Topic Selection | creator-buddy | 1.6k | Cross-platform viral and pain point finder |
Benchmarking | blogger-distiller | 694 | Distill blogger's viral tactics |
Layout/Images | baoyu-skills | 26.4k | Covers, illustrations, WeChat publishing |
Cards | guizang-social-card-skill | 7.4k | Xiaohongshu sets & WeChat cover pairs |
Visuals | Punk-Skill | 1.2k | Cover & avatar images (Commercial needs auth) |
Illustrations | ian-xiaohei-illustrations | 12.4k | Xiaohei-style body illustrations |
Formatting | write-then-publish | 561 | One text for cards and long-form |
Subtitle Summary | video-to-subtitle-summary-skill | 213 | Long video to subtitle + summary |
Storyboard | seedance-prompt-skill | 2.9k | Output Seedance 2.0 prompts |
Talking Head | video-talkcraft | 1.4k | Word-sync commentary video (Non-commercial) |
Video Distribution | video-publisher-skill | 229 | Five-platform drafts, default no publish |
Xiaohongshu | content-boom-monitor | 106 | Keyword & benchmark account monitoring |
Xiaohongshu | xiaohongshu-ops-skill | 2.4k | Full-chain operations & replication |
wewrite | 3.4k | Run whole pipeline with one sentence | |
Wechatsync | 6.4k | One-click sync to 29+ platforms | |
Douyin | jianying-headless | 3.0k | JianYing draft automation (macOS only) |
Douyin | yichen-skills | 4.3k | Writing, slicing, JianYing (Non-commercial) |
Twitter/X | x-article-publisher | 5 | Markdown to X Articles draft, no auto-publish |
Optional | hyperframes | 58.9k | Write HTML, render video |
Optional | remotion | 62.5k | React video framework core |
Optional | remotion-video-creation | 275k | React video, 29 rules |
Optional | manim-video | 275k | Code-drawn explanatory animations |
Optional | video-use | 28.4k | Chat-edit, auto-remove fillers/color grade |
Optional | jianying-editor-skill | 3.8k | JianYing draft automation (Export limited to Windows) |
Final Thoughts
Tools are fixed; rhythm is flexible. The same topic may yield different results on Xiaohongshu vs. WeChat—Skills save your hands, but not your judgment on where and when to publish.
Reference: <https://x.com/CrazyKaomei/status/2101208117304262843>
I am Xiangfeng @xaiwind, a creator focused on going overseas, AI, and AIGC. #SelfMedia #Skill
This article was created with AI assistance and personal practice. If helpful, please like, save, and follow.
I maintain the article continuously. Contact me or see you in the comments. Progress a little every day, and we will eventually reach our destination!
