claudeers.
// Other

video-agent-skills

Free Claude Code skills to script, research and edit short-form videos. Markdown only.

// Other[ cli ][ api ][ claude ]#claude#other◷ MIT$open-sourceupdated about 1 month ago
Actively maintained
98/100
last commit 15 days ago
last release none
releases 0
open issues 1
// star history

Install with your AI

Paste into Claude Code, Cursor, or any agent — it reads the repo and wires the tool into your project.

Install and set up video-agent-skills (git-clone project) into my current project.
Found on https://claudeers.com/video-agent-skills
Repo: https://github.com/Melvynx/video-agent-skills
Homepage/docs: —
Detected install method: git-clone → git clone https://github.com/Melvynx/video-agent-skills
Category: other. Platforms: cli, api.
Read the repo's README for exact setup and env vars, then install it and wire it into my project.

Claudeers Health Verdict:
active; community-verified: false. Confirm the source before running anything.
// or clone
git clone https://github.com/Melvynx/video-agent-skills

// compatibility

Platformscli, api
Operating systems—
AI compatibilityclaude
LicenseMIT
Pricingopen-source
Language—

Get your FREE $2.50 API credits to access TickAtlas financial data ↗

video-agent-skills

Free Claude Code skills to script, edit, translate and publish your videos: short-form (YouTube Shorts, TikTok, Instagram Reels) and full YouTube videos.

Only markdown. No app, no server. The skills tell Claude how to work and which tools to call: ffmpeg, yt-dlp, and Remotion with whisper.cpp for the edit. Optional: Treg to audit your Instagram and publish on YouTube, Instagram and TikTok, an ElevenLabs key to translate, Codex to generate images.

The skills

SkillWhat it does
/video-setupStart here. Sets up video editing on your computer and in your project: checks the tools, creates video-agent/, prepares the editing studio, writes your voice.md from your own videos or a short interview, pre-approves the commands, and optionally connects Treg, ElevenLabs and Codex.
/video-auditInstalls Treg if needed, connects your Instagram and audits it: your last reels with real stats (views, watch time, saves, shares), what over-performs against your own median, the first seconds of your best and worst reels, and the 3 changes to make, in audit.md.
/video-viralFinds what over-performs in your niche: recent Shorts of 10 to 12 accounts near your size and above, outliers against each account's median, the first 3 seconds of each, saved to swipe.md.
/video-scriptA week of scripts from what really happened in your week, in your voice: 3 hooks per script from 22 formulas, scored, a re-hook, timing checked at your pace, visual notes, then a humanize pass (invisible characters, AI words in English and French, AI sentence shapes, facts checked). Also hook <topic> and humanize <text> for any caption or description.
/video-editA raw take into a finished video: a short (1080×1920, captions in the platforms' safe zone, hook card) or a full YouTube video (1920×1080, chapters). Word-by-word captions, silences and retakes cut, simple motion design (callouts, punch-in zooms), voice at -14 LUFS, an .srt for CapCut or Premiere, and the post copy.
/video-stealCopies another creator's editing style from one of their reels: measures it with ffmpeg (cuts per second, layouts, captions, the beat and the sound effects on a spectrogram, the music level under the voice, the face size and the zooms), writes it to video-agent/styles/<name>.md, then edits a test video with video-edit and compares the two frame by frame. Then /video-edit <take> style <name> uses it.
/long-to-short-videosOne long video (yours, a podcast, a live) into finished shorts: the moments that stand alone, each one edited with video-edit (captions, hook card, callouts), plus new scripts from the ideas too spread out to clip.
/video-translateA take in another language with your own cloned voice, through ElevenLabs Dubbing and your key, then edited with captions in the new language.
/codex-imagesThumbnails, covers and stills generated with Codex's image tool (your ChatGPT plan), several in parallel, with your photos as reference.
/video-publishPublishes a finished video on YouTube (Shorts or long, scheduled or now), Instagram Reels and TikTok through Treg, with the light export, the cover, the caption and the title. Nothing goes out before your yes. Without Treg, a ready-to-post kit.

Install

In your project folder:

npx skills add Melvynx/video-agent-skills -a claude-code

Add -g to install them for every project. npx skills update gets the latest version.

Or paste https://github.com/Melvynx/video-agent-skills into Claude Code and ask it to install these skills.

Or copy them by hand (into ~/.claude/skills/ instead for every project). macOS and Linux:

git clone https://github.com/Melvynx/video-agent-skills.git
mkdir -p .claude/skills
cp -r video-agent-skills/skills/* .claude/skills/

Windows (PowerShell):

git clone https://github.com/Melvynx/video-agent-skills.git
New-Item -ItemType Directory -Force .claude\skills | Out-Null
Copy-Item -Recurse -Force video-agent-skills\skills\* .claude\skills\

To update a manual install: git -C video-agent-skills pull, then copy again.

Then, in Claude Code: /video-setup.

On claude.ai: zip one skill folder and upload it in the Skills part of the settings. Only video-script works there (scripts, hooks, humanize); the others need your computer's tools.

The flow

/video-setup           once: tools, studio, voice.md, permissions, accounts
/video-audit           what works on your own Instagram → audit.md
/video-viral           what works in your niche → swipe.md
/video-script          this week's scripts → video-agent/scripts/
                       record them (phone, camera, webcam) → video-agent/takes/
/video-steal           a creator's reel → a style video-edit can use
/video-edit            each take → video-agent/edits/<slug>/final.mp4 (short or long)
/long-to-short-videos  a long video → finished shorts
/video-translate       a take → the same take in another language, your voice
/codex-images          thumbnails and covers
/video-publish         YouTube, Instagram, TikTok

Everything lives in one folder at the root of your project:

video-agent/
├── voice.md      who you are and how you talk
├── swipe.md      hooks that worked in your niche
├── audit.md      what works on your own Instagram
├── log.md        one line per script, so nothing repeats
├── scripts/      one file per week
├── takes/        raw recordings
├── styles/       styles copied with video-steal (<name>.md + its stills, sfx, stickers)
├── edits/        one folder per edited video (final.mp4, post.md, cover.jpg, published.md)
├── repurpose/    one plan per long video
├── images/       generated images not tied to a video
├── studio/       the Remotion project
├── .env          your API keys (git-ignored)
└── .tmp/         scratch files

Tools

ToolForCost
Node 18+, gitthe edit (Remotion)free
ffmpegcuts, audio, conversionsfree
yt-dlp + denoYouTube views, captions, transcriptsfree
build toolswhisper.cpp compiles on the first edit (xcode-select --install, build-essential)free
Remotion + whisper.cppcaptions and motion design, downloaded on the first editfree for individuals and small teams (Remotion license)
Treg (optional)your Instagram audit, publishing on YouTube, Instagram and TikTok, TikTok and Instagram numbers of other accounts (CLI with Python 3.12+ on macOS and Linux, or the MCP on any OS)your own accounts free, other accounts about $0.001 a call, $1 free credit on signup
ElevenLabs (optional)video-translate, with your API keyyour ElevenLabs plan
Codex (optional)codex-images (npm i -g @openai/codex)your ChatGPT plan

video-setup gives the install command for each, on macOS, Windows and Linux.

Permissions

video-setup merges a short allow list into .claude/settings.json (yt-dlp, ffmpeg, ffprobe, npx remotion, node sub.mjs, and Treg's free read-only commands). Never Bash(*). Paid Treg calls always ask first, and the skills say the price before. Publishing and dubbing wait for your explicit yes. Keys live in video-agent/.env, never in a chat or a commit.

Honest limits

  • The edit is simple on purpose: captions, cuts, a hook card, callouts and zooms. No b-roll, music only from a file you bring.
  • Auto-captions mishear names and numbers. Check them.
  • Without Treg, research is YouTube only and publishing is manual (the skill hands you a ready kit).
  • Translation has no lip-sync: the image stays the original.
  • Generated images need a check: text can come out misspelled.
  • Only download and cut videos you own or have the rights to.

Want the full version?

These skills are the free base of Editing IA Pro: the full motion design engine, styles, music, a review studio, faceless avatars, thumbnails and titles.

License

MIT

// faq

What is video-agent-skills?

Free Claude Code skills to script, research and edit short-form videos. Markdown only.. It is open-source on GitHub.

Is video-agent-skills free to use?

video-agent-skills is open-source under the MIT license, so it is free to use.

What category does video-agent-skills belong to?

video-agent-skills is listed under other in the Claudeers registry of Claude-compatible tools.

46 views
★ 156 stars
unclaimed
updated about 1 month ago

// embed badge

video-agent-skills on Claudeers
[![Claudeers](https://claudeers.com/api/badge/video-agent-skills.svg)](https://claudeers.com/video-agent-skills)

// retro hit counter

video-agent-skills hit counter
[![Hits](https://claudeers.com/api/counter/video-agent-skills.svg)](https://claudeers.com/video-agent-skills)

// reviews

// guestbook

0/500

// related in Other

🔓

符合nature论文学术表达和科研绘图的Skill

// otherYuan1z0825/⟨Python⟩★ 54,085◷ Apache-2.0[ claude ]
🔓

Anti-AI-slop design skill for Claude Code, Cursor, and Codex.

// otherNutlope/⟨CSS⟩★ 49,642◷ MIT[ claude ]
🔓

Huashu Design · HTML-native design skill for Claude Code · Claude Code 里 HTML 原生的设计 skill · 高保真原型 / 幻灯片 / 动画 + 20 设计哲学 + 5 维评审 + MP4 导出 · Agent-agnostic

// otheralchaincyf/⟨HTML⟩★ 28,944◷ MIT[ claude ]
🔓

Open source Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents. Built for multitasking, organization, and programmability.

// othermanaflow-ai/⟨Swift⟩★ 28,842◷ NOASSERTION[ claude ]
→ see how video-agent-skills connects across the ecosystem