Skip to content
Get packs

Guide · Agent skill

How to use the video-editing agent skill: setup, commands and examples

Updated · written with AI from the project's own documentation and checked against our catalogue data

The video-editing skill is an agent skill that teaches an AI agent, such as Claude Code or Codex, a six-layer workflow for editing real footage that you already recorded. It is built for cutting long recordings, structuring vlogs and tutorials, and reframing video for different platforms, not for generating videos from text prompts. After this guide you will know what the skill does, how to install it, how to phrase requests so it activates, and which safety checks to run before you execute its commands.

Video Editing Workflowsaffaan-m/ECC · MIT

Quick start

  1. 1Download and unzip. Download the video-editing ZIP from this page and unzip it. Confirm the folder contains SKILL.md.
  2. 2Place the skill folder. Move the video-editing folder to ~/.claude/skills/ for personal use or to .claude/skills/ inside a project. Alternatively upload the ZIP under Settings → Capabilities → Skills.
  3. 3Install FFmpeg. Install FFmpeg locally, since the cutting, proxy and normalization steps depend on it. Add Node.js and Remotion only if you need overlays or compositions.
  4. 4Prepare your footage. Keep raw files in one folder and make a working copy. Create a transcript or timestamp list for the agent to plan from.
  5. 5Ask for a plan first. Describe the goal, target length and source file. Ask the agent to propose segments and cut commands before it runs anything.
  6. 6Review and run commands. Read each FFmpeg or Remotion command, confirm input and output file names, then run them step by step.
  7. 7Finish in an editor. Import the assembled video into Descript or CapCut for pacing, captions, color, audio mix and export.

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when a task matches the skill's description. This skill is named video-editing and is described as AI-assisted video editing workflows for cutting, structuring, and augmenting real footage.

The skill follows a pipeline: capture, organization, deterministic cuts with FFmpeg, programmable composition with Remotion, generated assets from ElevenLabs or fal.ai, and final polish in Descript or CapCut. The agent helps transcribe and plan cuts, writes FFmpeg commands to trim, join, proxy and normalize audio, and suggests Remotion code for overlays and motion graphics. The final polish is a human step.

  • Plans cuts from transcripts and identifies dead sections, pauses and repeated takes
  • Writes FFmpeg commands for segment extraction, concatenation, proxies, audio extraction and loudness normalization
  • Scaffolds Remotion compositions for titles, overlays, data visuals and product demos
  • Optionally calls ElevenLabs for voiceover and fal.ai for music, sound effects or b-roll

Install the video-editing skill

Download the skill ZIP from the item page on this directory and unzip it. For a personal setup, place the video-editing folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder. If you use the Claude apps, you can instead upload the ZIP under Settings → Capabilities → Skills.

The download includes agents/openai.yaml and SKILL.md. The SKILL.md file is the part the agent reads for instructions.

  • Personal install: ~/.claude/skills/video-editing/
  • Project install: .claude/skills/video-editing/ inside your project
  • Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Install the tools the workflow needs

The skill calls local tools rather than doing video work itself, so you need them installed before the agent can run its commands. FFmpeg is required for cutting and reframing. Node.js and Remotion are needed for overlay and composition steps.

Voice, music and generation services are optional. Only set them up for the layers you plan to use. The source does not provide install commands for these tools, so follow each tool's own documentation.

  • Required: FFmpeg installed locally
  • Required for overlays and compositions: Node.js and Remotion
  • Optional: ElevenLabs API key in the ELEVENLABS_API_KEY environment variable, for voiceover
  • Optional: fal.ai account and the fal-ai-media skill, for music, sound effects and generated images
  • Optional: VideoDB account and configuration, for desktop capture and generative audio
  • Human step: Descript or CapCut for final polish

How to trigger the skill with natural requests

The skill activates when your request matches its description. Plain language works well. The source lists phrases such as "edit video", "cut this footage", "make a vlog" and "video workflow". Give the agent the file, the goal and the length you want so it can plan cuts.

A useful request names the source material, the target length and the structure you want. The skill is most useful when the agent first reads a transcript and proposes segments before writing any commands.

  • "Edit video: here is the transcript of a 4-hour recording. Pick the 8 strongest segments for a 24-minute vlog and give me FFmpeg cut commands."
  • "Cut this footage into segments from my timestamp list and join them in order."
  • "Reframe this 16:9 demo for TikTok and Instagram Reels."
  • "Add a title overlay and lower third to this tutorial using Remotion."
Example prompt: "Here's the transcript of a 4-hour recording. Identify the 8 strongest segments for a 24-minute vlog. Give me FFmpeg cut commands for each segment."

Example commands the workflow produces

These are examples from the skill source. Check the input file names and timestamps before running them, because the FFmpeg cut and concat steps write output files.

To extract a segment by timestamp, the skill uses stream copy, which avoids re-encoding. To make a lightweight proxy for faster editing, it scales the video to 960 pixels wide.

ffmpeg -i raw.mp4 -ss 00:12:30 -to 00:15:45 -c copy segment_01.mp4 ffmpeg -i raw.mp4 -vf "scale=960:-2" -c:v libx264 -preset ultrafast -crf 28 proxy.mp4 ffmpeg -i raw.mp4 -vn -acodec pcm_s16le -ar 16000 audio.wav ffmpeg -i segment.mp4 -af loudnorm=I=-16:TP=-1.5:LRA=11 -c:v copy normalized.mp4

Reframe video for YouTube, TikTok and Instagram

The skill maps common platforms to aspect ratios: YouTube at 16:9 (1920x1080), TikTok and Reels at 9:16 (1080x1920), Instagram Feed at 1:1 (1080x1080), and X or Twitter at 16:9 or 1:1 (1280x720 or 720x720). For vertical or square versions, it suggests center-crop FFmpeg commands.

  • Ask for the target platform by name so the agent picks the right ratio
  • Check the crop: center crop can cut off faces or subjects near the edges
ffmpeg -i input.mp4 -vf "crop=ih*9/16:ih,scale=1080:1920" vertical.mp4 ffmpeg -i input.mp4 -vf "crop=ih:ih,scale=1080:1080" square.mp4

Safety notes before you run anything

This skill runs scripts and uses the network. The FFmpeg and Remotion steps can overwrite or misplace files if run without checking, so read each command before you execute it. Keep your original raw footage in a separate folder and work on copies.

Network calls happen only when you use optional services. The voiceover example sends your text to ElevenLabs using your API key, and the fal.ai and VideoDB steps call their services. Do not paste secrets into prompts, and keep API keys in environment variables.

  • Review every command before running it, especially batch scripts and concat steps
  • Back up raw footage and write outputs to new file names
  • Keep API keys in environment variables such as ELEVENLABS_API_KEY
  • Remember that generated audio and visuals are sent to third-party services when you use them

What people use it for

  • Cutting a 4-hour recording into a 24-minute vlog using timestamped segments
  • Turning a product demo screen capture into a short tutorial with Remotion title overlays
  • Reframing one 16:9 video into 9:16 for TikTok and Reels and 1:1 for Instagram
  • Creating a voiceover track with ElevenLabs for an existing narrated clip
  • Normalizing audio levels across segments recorded on different days

Frequently asked questions

Is the video-editing skill free?

The skill is published under the MIT license, as stated on this page. The tools it uses have their own terms: FFmpeg and Remotion are installed locally, while ElevenLabs, fal.ai and VideoDB are optional accounts that may have their own costs.

What do I need to use the video-editing skill?

You need FFmpeg installed locally for cutting and reframing. Node.js and Remotion are needed for overlay and composition steps. ElevenLabs, fal.ai and VideoDB are optional and only needed for voiceover, music, sound effects, generated images or desktop capture.

Can this skill generate a full video from a text prompt?

No. The skill is designed for editing existing footage, not generating whole videos from prompts. Its approach is to use AI to compress, structure and augment real recordings.

Where do I install the video-editing skill?

For personal use, place the folder in ~/.claude/skills/. For one project, use the .claude/skills/ folder in that project. In the Claude apps, upload the ZIP under Settings → Capabilities → Skills.

Why does the agent write FFmpeg commands instead of editing directly?

The skill uses FFmpeg for deterministic cuts like trimming, joining and normalizing, which are repeatable from the command itself. Its guidance is to keep the agent focused on structure and let each tool handle its own job.

Is it safe to run the commands the agent writes?

Read them first. The source warns that FFmpeg and Remotion steps can overwrite or misplace files if run without checking, and the skill runs scripts and uses the network. Work on copies and use new output file names.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits creators turning long recordings into shorter videos, developers making demos or tutorials from screen captures, and marketers reframing footage for YouTube, TikTok and Instagram. It is not the right choice if you want a video generated from a text prompt, or if you are not comfortable reading and checking the FFmpeg and Remotion commands before running them.

Download the skill