Video editing skills for your AI agent

Choose the workflow that fits your footage

Install the ChatGPT and Codex plugin

The plugin combines the editor connection with four skills: get-started helps you connect and select a recording; the other three guide footage editing, shorts and website demos. Review the versioned release and downloadable package, then use a supported Codex CLI to add Valmera's repository marketplace:

codex plugin marketplace add ABO3SKRALMASAOODI/valmera-mcp --ref openai-v1.0.0
codex plugin list --marketplace valmera --available --json

1. Find the package. The list command should show valmera@valmera version 1.0.0. This means the package is available, not installed or authenticated.
2. Install and connect. In the desktop Plugins Directory, select Valmera video editing, install Valmera and complete the Valmera connection setup. Client support and workspace policies can affect availability.
3. Choose your recording. Run setup or ask the assistant to list your Valmera projects. Select the intended recording, or upload one in Studio, before requesting an edit.

This package is distributed from Valmera's repository. OpenAI public Directory submission and approval have not been completed. For ChatGPT's custom MCP connection route, follow the ChatGPT setup guide; that route connects the editing tools separately from these bundled skills. Avoid adding a duplicate connection if you already have one.

Installing the plugin does not upload footage or start a paid edit. Account creation and uploads are free; editing requires a paid Valmera plan. Your assistant access and usage are separate.

Marketplace fetch and discovery were verified with Codex CLI 0.166.0-alpha.5.1. The downloaded ZIP matched the published source; authenticated installation and a paid editing run were not tested. See OpenAI's plugin and marketplace instructions for client setup.

Install a skill in your project

Use Node.js 22.20 or later. Open a terminal in the project where you want the skill, review its linked source, then run the command for your agent. These commands pin both the Skills CLI and Valmera release.

npx skills@1.7.6 add ABO3SKRALMASAOODI/valmera-mcp@v1.0.1 --skill edit-video-with-valmera -a claude-code

npx skills@1.7.6 add ABO3SKRALMASAOODI/valmera-mcp@v1.0.1 --skill edit-video-with-valmera -a codex

npx skills@1.7.6 add ABO3SKRALMASAOODI/valmera-mcp@v1.0.1 --skill edit-video-with-valmera -a cursor

For another workflow, replace edit-video-with-valmera after --skill with turn-long-video-into-shorts or website-demo-video. Follow the installer prompts, then start or refresh your agent session so it can discover the skill.

For Claude chat, Cowork or Claude Code, the Claude video editing plugin bundles the same three skills with the editor connection. The guide includes a standalone ZIP and marketplace setup. Inspect your installed version before adding another copy.

For Gemini CLI, follow the Valmera extension guide to install the connection and editing context together.

Next, connect and authenticate Valmera at https://valmera.io/mcp/server. Ask the assistant to list your projects. An empty list can be a valid connection result; choose or upload the recording you intend to edit.

Start with a brief the agent can act on

Choose one starting prompt below and replace its project or URL reference with your own. Name any required caption language, brand assets, music and delivery files.

Use edit-video-with-valmera on the Valmera project I select. Make a 45-second vertical edit for Instagram from my uploaded recording. Preserve the speaker's meaning, keep names accurate and use readable captions. Confirm the project and cost expectations before billable work. Show me a preview before final export.

Use turn-long-video-into-shorts on my selected podcast project. Propose three self-contained stories for 30–60 second vertical clips. Show the source ranges and why each stands alone. Keep the active speaker in frame and confirm caption requirements and costs. Review each preview before exporting the clips I approve.

Use website-demo-video to plan a 30-second walkthrough of the public website URL I provide. Show the product, one useful action and its result. Use only the pages and actions I approve; keep private account details out of the capture. Confirm the capture route, voice or music requirements and cost before execution, then show me a preview.

1. Choose the source. Confirm the project and wait for its analysis to finish. A local path mentioned in chat does not upload a file; follow the upload guide.
2. Agree on the first pass. Set the destination, aspect ratio, target length and cost expectations. If the account does not expose an exact quote, approve a bounded preview before billable work.
3. Review the video. Check meaning at cuts, caption spelling, subject framing and sound. When requesting a revision, say whether its timestamp refers to the original source or the shortened preview.
4. Deliver the reviewed version. Confirm the project and edit version before final export. Follow the existing job to completion and retrieve the final MP4 link; inspect the file against your brief.

When installation or editing stalls

For sign-in errors, missing tools and authentication challenges, use the connection troubleshooting guide. Before upgrading a locally customized skill, preserve your edits and review the new source.

Source and verification

The three skill files passed format validation. Release-tag installs were checked for Claude Code, Codex and Cursor in isolated projects; a commit-pinned editor-skill install was checked too. Installed files matched the published source. These checks verify installation and instructions, not a new paid editing run in every client.

The original trim-and-reframe example includes source ranges and a downloadable cuts-and-crop export. It demonstrates that recorded workflow, rather than every treatment described here.

Release: v1.0.1 · Source commit 61812b0 · Skills CLI source