Guide · Agent skill
How to use the Shotcut Video Editing CLI agent skill
Updated · written with AI from the project's own documentation and checked against our catalogue data
This guide explains the cli-anything-shotcut agent skill, which lets an AI agent create and change Shotcut video projects by running a command-line tool instead of using the editor's windows and buttons. It is written for readers who have never used agent skills before. By the end you will know what the skill does, how to install it, how to ask for work in plain language, and what it needs on your computer.
Shotcut Video Editing CLIHKUDS/CLI-Anything · Apache-2.0 Download the skill Quick start
- 1Install prerequisites. Make sure Python 3.10 or newer, Shotcut, melt, ffmpeg and ffprobe are installed on your system.
- 2Install the package. Run pip install cli-anything-shotcut in your terminal.
- 3Place the skill folder. Unzip the downloaded skill into ~/.claude/skills/ or a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
- 4Create a project. Ask the agent to create a project, or run cli-anything-shotcut project new -o project.json yourself.
- 5Import media. Import your files with media import so each gets a clip_id.
- 6Edit the timeline. Place, trim and split clips and add filters or ducking, then check the result with info or show.
- 7Render. Render the project to a video file with render and a preset, after reviewing the result.
What an agent skill is and what this one does
An agent skill is a folder containing a SKILL.md file that tells an AI agent when and how to use a tool. This skill's SKILL.md describes a stateful command-line interface for video editing, built on the MLT XML format and designed for AI agents and power users who want to create and edit Shotcut projects without a GUI.
Once installed, the agent can create projects, import media, place clips on a timeline, trim and split them, add filters and transitions, add audio ducking, and render the result to a video file. Changes can be undone and redone, and output can be returned as JSON for the agent to read.
- Create a project and import media files into its bin
- Place, trim, split and remove clips on video and audio tracks
- Add filters, transitions, blend modes and opacity settings
- Build volume envelopes and ducking envelopes for audio
- Render the project to a video file using export presets
Install the skill from this page
Download the ZIP for this skill from this page and unzip it. The skill folder contains SKILL.md. Put the folder in one of these locations: ~/.claude/skills/ for all your projects, or a project's .claude/skills/ folder for one project only.
If you use the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.
- Personal install: unzip into ~/.claude/skills/
- Project install: unzip into the project's .claude/skills/ folder
- Claude apps: Settings → Capabilities → Skills, then upload
Install the command-line tool and its dependencies
The skill drives a command-line tool that is installed as a Python package. The SKILL.md gives this install command. You also need the other programs listed below, because the tool depends on them.
Prerequisites: Python 3.10 or newer; melt (the MLT command-line tool), needed for rendering and playback; ffmpeg and ffprobe, needed for media probing; and Shotcut itself installed on the system.
- Python 3.10+
- melt (MLT CLI) for rendering and playback
- ffmpeg and ffprobe for media probing
- Shotcut installed on your system
pip install cli-anything-shotcutAsk for edits in natural language
You do not need to memorise commands. Describe the result you want, and the agent chooses the commands. The source notes one workflow rule the agent follows: import media first to get a clip_id, then place that clip on the timeline with add-clip.
Be specific about the project file, the media files, the order of clips and the output file. Ask the agent to report the project info or the timeline so you can check its work before rendering.
- Example request: Create a new 1080p project called interview.mlt, import clip1.mp4 and clip2.mp4, and place them one after the other on track one.
- Example request: Trim the first clip so it starts at 00:00:02.000, then render the project to rough_cut.mp4 with the default preset.
- Example request: Add a ducking envelope to the music track so it lowers during the dialogue between 00:00:06.000 and 00:00:09.000.
Commands the agent can run
The agent can run the tool directly from a terminal. These are the main commands taken from the source. Adding --json before a command asks for JSON output, which the agent can read.
Undo and redo are available for the operation history, and the interactive mode starts when the command is run without a subcommand.
- cli-anything-shotcut --help shows help
- cli-anything-shotcut project new -o project.json creates a new project
- cli-anything-shotcut --json project info -p project.json returns project info as JSON
- media import <file> returns a clip_id such as clip0
- add-clip <clip_id> <track> places the clip on a track
- render <output> --preset name renders to a video file
cli-anything-shotcut project new -o project.jsonSafety notes before you run it
The skill runs scripts on your computer. Its listed requirements say it runs scripts and that it does not use the network. The tool reads and writes local media files and project files on the machine it runs on, so do not point it at sensitive footage on a shared computer without thinking about who can access those files.
Review what the agent did before you render or share the result. The agent cannot judge creative quality, so a finished cut still needs a human check.
- It runs scripts on the machine where the agent runs
- It reads and writes local media and project files
- It makes no network use according to its listed requirements
- Check the project info or timeline before you render
Tips for better results
Keep media file paths simple and confirm they exist, since the tool has a check command for media files. Use timecodes in the hours:minutes:seconds.milliseconds form shown in the source, such as 00:00:03.000. Save the project often, and use undo if an edit goes wrong.
For previews during iterative review, the source describes preview bundles and a live preview session. Those are optional and need the viewer commands from the same toolkit.
- Use timecodes like 00:00:03.000
- Run the media check command before building a timeline
- Save with the save command so you keep a project file to return to
- Use the preview commands to review low-resolution bundles before a full render
What people use it for
- Assembling a rough cut from a set of interview clips on a timeline
- Repeating the same trims and filters across many projects from a script
- Building an automated video production pipeline that outputs rendered files
- Adding ducking to background music under dialogue in a project
Frequently asked questions
Is the Shotcut agent skill free?
The source file does not state a licence or price for the skill beyond the Apache-2.0 licence shown on this page. Check the repository licence for the terms of use.
Do I need Shotcut installed to use it?
Yes. The listed prerequisites say Shotcut must be installed on your system, along with melt, ffmpeg and ffprobe. Rendering and media probing depend on those programs.
What do I need to install first?
You need Python 3.10 or newer, the pip package cli-anything-shotcut, melt for rendering and playback, ffmpeg and ffprobe for media probing, and Shotcut itself.
Does the skill need an internet connection?
The listed requirements say it does not use the network. It works on the local media and project files on your computer.
Why does the agent import media before adding clips?
The source sets a two-step model. Media import returns a clip_id, such as clip0, and add-clip uses that id to place the clip on a track.
Can the agent check its own edits?
It can read project info, the timeline and JSON output, and it can undo or redo steps. It cannot judge creative quality, so review the result yourself before you use the render.
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 who want an agent to assemble rough cuts or repetitive edits in Shotcut, and developers who want a scriptable editing interface for pipelines. Skip it if you need a polished, finished edit without reviewing the output, or if you cannot install Shotcut, melt, ffmpeg and ffprobe. Be careful with sensitive footage on shared machines, since the tool reads and writes local files.
Download the skill