by PansaLegrand
Open-source video editor where you and an AI agent work on the same timeline. Run locally, bring your own model keys. Powered by Dojo Video Engine.
Available after the catalog security scan passes.
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
The deep catalog scan for this skill is still queued. Run an instant dependency check now instead.
# Add to your Claude Code skills
git clone https://github.com/PansaLegrand/dojo-video-agentGuides for using ai agents skills like dojo-video-agent.
See how dojo-video-agent compares with popular alternatives.
An open-source video editor where you and an AI agent work on the same timeline. Run locally, bring your own model keys, and extend the TypeScript agent. Import footage, edit by hand or through chat, and export an MP4 in your browser. Ordinary validated chat edits apply directly; Compare and Undo let you inspect or reverse them.
Want to try the hosted product? Open DojoClip Video Edit AI. DojoClip adds managed accounts, projects, provider jobs, storage, and payments. This repository is the smaller local editor and developer foundation; it does not install the commercial SaaS.
https://github.com/user-attachments/assets/99fa44aa-1e46-4594-9938-7b8cfe312223
Illustrated stories, animated data, portrait typography, cinematic titles and 3D branding.
https://github.com/user-attachments/assets/59f7b7d2-bb9c-442a-808c-8935aeec4de0
A miniature city, an exploded mechanical timepiece and layered kinetic vector art.
https://github.com/user-attachments/assets/088f91fb-d8a0-47d6-85c9-9e7665a116e9
Sports broadcast graphics, a software workflow and a colorful drink campaign.
Explore the editable examples and make them yours → The original scenes stay editable; no API keys are needed to render them.
Formerly Dojo Video Harness (dojo-video-harness). The repository name changed; package names, saved projects, and runtime profiles remain compatible. See the rename record.
| Project | Use it for | What you operate |
|---|---|---|
| Dojo Video Agent | Edit local footage, study the agent, add tools or skills | A single-user localhost app; your own optional provider keys |
| Dojo Video Engine | Build scene rendering, playback, inspection, and export into your own application | TypeScript packages and browser media APIs |
| DojoClip | Use the hosted editor and managed AI services | An account in the commercial product; no local server setup |
Dojo Video Agent includes the shared editor UI, command validation, timeline history, captions, editable graphics, browser preview/export, and a bounded model/tool loop. It deliberately excludes commercial authentication, billing, Postgres/Prisma, Redis queues, cloud-storage configuration, production workers, and deployment credentials. Local image/video-generation adapters are not shipped; the optional BYOK music adapter is available.
No DojoClip database or commercial environment file is needed. Projects and receipts use local files; imported originals use browser IndexedDB. The UI is English-only, while media text and provider-supported captions can use other languages. See the scope and synchronization policy.
Requirements: Git, Node 20.19+ (Node 22 recommended), npm, and a desktop browser with WebCodecs support. Chromium is the tested path. FFmpeg is only needed for acceptance tests.
git clone https://github.com/PansaLegrand/dojo-video-agent.git
cd dojo-video-agent
npm run launch
launch installs locked dependencies when missing, creates .env without overwriting an existing file, checks the environment, and starts the editor. It opens a browser from an interactive terminal; otherwise open http://127.0.0.1:4310. The published engine packages install automatically. No sibling repository or pnpm installation is needed.
Create a project and import a short clip. Manual editing, subtitle import, preview, and MP4 export need no API keys. For chat or speech recognition, open Provider settings and configure the relevant provider below.
npm run dev # Start again after initial setup
npm run setup -- --install # Install updated locked dependencies after git pull
npm run doctor # Check Node, local storage, and port availability
npm run dev -- --no-open # Start without opening the browser
The launcher supports PowerShell, macOS, and Linux. Keep the same local address: browser media belongs to its origin. An occupied port produces guidance rather than silently switching addresses. To run a production build locally, use npm run build then npm start.
With the local editor running, use another terminal at the repository root:
node examples/creative-showcase/import.mjs journey
node examples/creative-showcase/import.mjs data-story
node examples/creative-showcase/import.mjs kinetic
node examples/creative-showcase/import.mjs cinema
Each command creates a fresh project and prints its editor link. Play it, select a graphic, make it yours and export. Source, customization and rendering details.

These examples use capabilities exercised by the repository's tests. Chat wording is a starting point: inspect the resulting edit, and use Undo or a follow-up instruction when needed. See verification for the distinction between scripted tests and live-model evidence.
This is a local editing workflow. It needs neither an AI provider nor an account. Rendering and export run in the browser; supported media depends on browser codecs and admission limits.
In an open project, choose Whole project scope and ask:
Add a title saying “Hello” for the first three seconds.
Then send each follow-up separately:
Make it red.
Put a yellow backing behind it.
Make the text bigger.
The title stays editable on the timeline. Ordinary local edits apply automatically; Compare and Undo remain available. The bounded correction diagnostic records a tested four-turn version of this workflow, including preserved object identity and rendered changes.
Import a short spoken clip, choose Whole project scope, and ask:
Add English captions to this clip, then add Spanish translations below them. Keep the original English captions.
If transcript evidence is missing, the app requests source/provider consent before transcription and then continues the original instruction. Review timing, wording, and translated facts before export. Transcription uploads are capped at 25 MiB and require genuine word timestamps. Already imported subtitle text can be edited manually without an ASR key.
For an explicit multi-clip workflow, import two spoken clips named Clip A and Clip B without existing linked captions, then ask:
Join Clip A then Clip B, generate English captions, translate them to Spanish, then export.
That workflow has its own review card: approve provider access, review and apply the prepared edit, then use Export approved workflow. The saved export is bound to the applied revision. See workflow behavior and evidence.
With footage on the timeline, ask:
Add a three-second 3–2–1 countdown before my footage. Move the footage after the countdown and keep its full duration.
Or:
Add a two-second animated “Thanks for watching” ending after my footage.
These can be native grouped graphics with editable text and animation. They do not require generating another video asset. You can move, trim, split, or edit the graphic and export it with the footage. Trimming cuts animation beats; it does not automatically retime them or ripple neighboring clips. See the intro/outro tests and limitations.
Configure Background music · Kie in /en/settings, then use the Background music panel or ask through chat. The panel works without a GPT key. Generate and insert music approves one paid job and its displayed placement, length, and volume. The first returned track is stored locally and inserted when ready. Undo removes an edit; it does not repeat generation or refund a provider charge. See music setup and recovery.
| Operation | Credentials |
|---|---|
| Manual edits, subtitle import, preview, MP4 export | None |
| Chat ed |
dojo-video-agent is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by PansaLegrand. Open-source video editor where you and an AI agent work on the same timeline. Run locally, bring your own model keys. Powered by Dojo Video Engine. It has 56 GitHub stars.
dojo-video-agent's catalog security scan is still queued. You can run an instant dependency and prompt-injection check now with the "Scan for vulnerabilities" button above.
Clone the repository with "git clone https://github.com/PansaLegrand/dojo-video-agent" and add it to your Claude Code skills directory (see the Installation section above).
dojo-video-agent is primarily written in TypeScript. It is open-source under PansaLegrand on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh dojo-video-agent against similar tools.
No comments yet. Be the first to share your thoughts!