skills
來自 scenario-labs/skills 的 67 項代理技能 — 包含 scenario、scenario-game-assets、scenario-video。使用 npx skills add scenario-labs/skills 安裝。
AI 翻譯
尚未偵測到更新。GitHub 發布版本後會顯示在這裡。
個人 GitHub 帳號下公開儲存庫的擁有者可以登入並驗證所有權,自行管理介紹,無需管理員核准。
登入並產生驗證檔案,在 30 分鐘內將其提交至此儲存庫的預設分支。其他分支或分叉儲存庫中的檔案無效。然後在下方檢查檔案,即可無需管理員核准直接管理專案。請保留檔案以維持管理權限。
如需登入,將前往登入頁面。登入後請回到此處,再次點擊所選的驗證按鈕。
在專案資料夾中執行 /nomorevibe verify,即可驗證所有權並自行管理介紹。
一分鐘完成驗證Create Instagram Reels, TikToks and YouTube Shorts with adaptive AI editing, captions and motion graphics. Connect FavStash for planning, publishing and analytics.
翻譯準備中
取自儲存庫 README
正在載入完整 README。 以下為已儲存的摘錄。
Give this prompt to your AI agent:
FavStash — AI Video Editing & Scripting Skills for Codex & Claude Code
Make your doom-scrolling usable. Save inspiration, create with your AI, review
your content and bring the results into your next plan.
Script and edit Instagram Reels, TikTok videos and YouTube Shorts with Codex,
Claude Code or another skills-aware AI agent. Start with a saved reference, a
topic or your footage. Let the agent write scroll-stopping hooks in your voice,
shape the cut, add readable captions, build useful motion graphics and prepare
a reviewed video export.檢視完整 README ↗Give this prompt to your agent:
Set up FavStash creator skills for my short-form workflow.
Clone https://github.com/alisadiq-ai/favstash-ai-video-editing-skills and follow:
https://raw.githubusercontent.com/alisadiq-ai/favstash-ai-video-editing-skills/main/INSTALL_FOR_AGENTS.md
Install all ten skills and the local media tools they need in my workspace.
Check whether FavStash is connected; if not, connect it with the FavStash CLI
(Codex or Claude Code) or guide me through sign-in. Verify the setup, then tell
me how to start my first reel.
Give your AI agent the whole short-form loop. These free, open-source skills teach Claude Code, Codex or any skills-aware agent to find proven ideas, write scroll-stopping scripts in your voice, edit Reels, TikToks and Shorts in five finished styles, then publish and learn from the results.
flowchart LR
R[1 · Research<br/>find-ideas] --> S[2 · Script<br/>write-script]
S --> E[3 · Edit<br/>edit-video + a style]
E --> P[4 · Publish & learn<br/>publish-and-analyze]
P -->|results pick the next idea| R
| Stage | Skill | What your agent does |
|---|---|---|
| 1 · Research | find-ideas | Finds proven reels to adapt (3× their creator's median), reads saved inspiration and suggests what to make next |
| 2 · Script | write-script | Copies the proven hook, adds as many equally aggressive alternatives as you want to test, writes one shared body in your voice and hands over a filmable brief |
| 3 · Edit | edit-video | Cuts your takes, removes retakes, adds captions, sound and motion graphics, keeps everything in safe areas and reviews the actual export |
| 4 · Publish & learn | publish-and-analyze | Exports at full quality, picks a real-frame cover, avoids duplicate holds on hook variants, schedules after your approval and compares results |
Every skill works on its own. FavStash closes the loop: your saved reels feed research, approved cuts go to Instagram, TikTok, YouTube or LinkedIn, and the analytics pick the next test.
edit-video is the base for every edit. Add the style that fits the story; each one
lives in its own folder under skills/3-editing-styles/.
For product, tool and explainer reels with real screen proof. Palettes are matched to your footage.
For tool, repo and feature explainers where every sentence names something to show.
For talking-head explainers that alternate graphics with personal lines. One-to-two-word captions in Montserrat.
For promoting a product, physical or digital (a gadget, a bottle, an app, a SaaS tool), a free resource or a comment-to-DM giveaway, with no camera footage. Your agent plans the story from the product itself (its app, site, photos or packaging), carries one object from scene to scene and locks every sound to its frame. It also makes a 16:9 launch film for YouTube or a landing page.
Ask for an effect by name: kinetic hook, gravity drop, giant captions, timeline, grade slider, title card, comment drop, orbit, frame break, vertigo, paper cut, rewind and 21 more. Big words sit behind you, and your own frames, posts and waveform appear inside the UI. Each effect has a copy-paste prompt.
For jokes, observations and setup/payoff reels, with footage matched to the line.
The skills are free and open source (Apache-2.0). FavStash's free plan includes unlimited saved items and stash searches, 100 transcript and AI enrichments, one connected channel and 10 posts a month. See the current plans.
Paste the prompt above into Claude Code, Codex, Cursor or another local coding agent. It follows INSTALL_FOR_AGENTS.md to install the skills, prepare FFmpeg, HyperFrames and GSAP in your workspace, render a smoke test and connect FavStash. You only sign in; the agent does the rest.
npx skills add https://github.com/alisadiq-ai/favstash-ai-video-editing-skills
Add --skill edit-video --skill paper-grid (for example) to pick skills. The
agent setup guide covers media tools and the FavStash connection.
For Claude Code or Codex, the FavStash CLI connects your stash and uploads local files:
npm install -g @sketric/favstash-mcp@latest
favstash auth login
favstash setup --agent claude-code
favstash doctor
Use --agent codex for Codex. Other agents connect through FavStash's MCP server
with OAuth (connection guide).
This repository is also a Claude plugin, favstash-creator-skills, with all ten
skills and the FavStash connector. Load it in Claude Code for one session:
claude --plugin-dir ./favstash-ai-video-editing-skills
Skills are namespaced under the plugin, for example
/favstash-creator-skills:edit-video, and Claude also picks them up when your
request fits. Connect FavStash from the plugin's connectors with OAuth. The editing
skills need the local media tools in the agent setup guide,
so use the plugin in Claude Code or in Cowork on your own computer.
Use $find-ideas to find three proven reels about AI coding tools I could adapt.
Use $write-script to adapt this reel into a few hooks to test and one body in my voice.
Use $edit-video with $breakout-card to edit these takes into a 40-second reel.
Use $edit-video with $product-motion-ads to make a 25-second ad for my app from its code,
ending on "comment TEMPLATE".
Use $edit-video with $motion-effects to add a kinetic hook, a timeline and a comment drop
ending on "comment EDITOR" to my talking-head video.
Use $publish-and-analyze to prepare covers and a caption for this cut, then compare
last week's hook test once I approve.
Do I need FavStash? No. Research, scripting and editing work without it. FavStash adds your saved inspiration, scheduling and analytics so the loop closes.
Will my agent post without asking? No. It delivers a local cut. Posting needs your approval for the exact file, cover, caption, account and time.
What runs where? Editing runs locally with FFmpeg and HyperFrames (Node 22+).
paper-grid, breakout-card and the duplicate check also use Python. The
breakout card's built-in person matte uses Apple Vision on macOS; on other systems,
supply a matte from any segmentation tool.
What does it send, and where? The bundled scripts run on your machine and make no
network requests; your footage, renders and preferences stay in your workspace.
Data leaves your machine only through FavStash tools you connect: the remote MCP
server at https://mcp.favstash.app/mcp (OAuth) or the FavStash CLI. They search
your stash, read your connected accounts' analytics, and upload and schedule a
post only after you approve the exact file, caption, account and time. See the
FavStash privacy policy.
Can I use a saved reel as a reference? Yes, for analysis. Adapt its hook, structure and pacing, and make the video from your own footage. Downloaded footage and music stay analysis-only unless you have reuse rights.
npm test
npm run validate
See contributing and security. Code, instructions and original SFX use Apache-2.0; external tools and the bundled Montserrat font keep their own licenses. The style previews come from the maintainer's own reels, with the speaker replaced by a placeholder.