How to use the Kdenlive video editing CLI agent skill

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

An agent skill is a folder with a SKILL.md file that teaches an AI agent how to perform a specific task with tools available on your computer. This skill teaches the agent to drive a stateful command-line interface for video editing that follows the patterns of the Blender CLI harness and uses a JSON project format with MLT XML generation for Kdenlive and melt.

The skill covers project management, a media bin, timeline tracks and clips, filters, transitions, guides, export, and session commands for undo, redo and history. The source asks the tool to return structured JSON so the agent can read results reliably.

• Create and open projects, and save or print project info
• Import, list and remove clips in the media bin
• Add, remove and move tracks and clips; trim and split clips at a time offset
• Add, set and remove filters and transitions
• Add, list and remove guides at a position in seconds
• Generate Kdenlive/MLT XML and list render presets
• Undo and redo, with up to 50 levels of history

Install the Python package and Kdenlive

The CLI is installed as the pip package cli-anything-kdenlive. You need Python 3.10 or newer and Kdenlive installed on your system. The skill does not install Kdenlive for you.

• Python 3.10 or newer
• Kdenlive installed on the system
• The pip package cli-anything-kdenlive

Add the skill to Claude

Download the skill ZIP from this page and unzip it. Then place the skill folder in one of two locations: ~/.claude/skills/ for personal use, or a project's .claude/skills/ folder if you want it limited to one project.

You can also upload the ZIP in the Claude apps under Settings → Capabilities → Skills. Choose whichever option matches how you use Claude.

• Personal: unzip into ~/.claude/skills/
• Single project: unzip into your project's .claude/skills/
• Claude apps: Settings → Capabilities → Skills, then upload the ZIP

Ask for editing work in plain language

You do not need to memorise commands. Describe the change you want, name the project file and the clip, and ask for a result. The agent chooses the matching cli-anything-kdenlive commands. Give absolute file paths where you can, because the source recommends them for all file operations.

If you want to check the tool yourself, the help command and the JSON flag are the starting points. The source describes --json as the mode for agent consumption.

• Name the project file, for example myproject.json
• Give the full path to each media file
• Say what should change: track, clip, in/out points, filter or marker
• Ask the agent to report the result as JSON or to show the project info afterwards

Example requests and the commands behind them

The agent will typically run a sequence of commands. The examples below show what the source documents for common tasks. Your request does not need to match these words exactly.

To start a new project, the source shows a project new command with an output flag. To render, it shows an export render example, but the source does not explain how rendering is performed, so do not assume a finished video will appear from that command alone.

• Request: 'Start a new project called interview.json.' Documented command: project new -o interview.json
• Request: 'Show me the project info as JSON.' Documented command: --json project info -p project.json
• Request: 'Add a guide at 12 seconds.' Documented command: guide add (position in seconds)
• Request: 'Undo the last change.' Documented command: session undo

Safety notes: scripts and network

The skill runs scripts, so treat it like any other code you install. The skill does not use the network according to its listing. Review the package's README.md before relying on it, since the source points there for full documentation.

The tool works on local media files. The documentation does not describe how file paths or data are handled, so avoid using it on sensitive or client footage until you have checked this yourself. Keep backups of original media and project files, and check return codes and stderr as the agent should do.

• The skill runs scripts; review them before installing
• The listing reports no network use
• Documentation does not describe data handling; avoid sensitive footage until you verify
• Keep backups of projects and originals

Tips for reliable results

Ask the agent to use JSON output so it can parse results and check return codes. Use undo and redo to back out of mistakes, and remember that history is limited to 50 levels. Verify that output files exist after an export operation, as the source advises.

Use the interactive REPL when you want to step through changes yourself. Running the CLI with no subcommand starts a session with tab-completion and history.

• Always use the --json flag for agent work
• Check return codes: 0 means success, non-zero means error
• Read stderr when a command fails
• Verify output files exist after export
• Use session history to undo mistakes