recut-editor

Edits Recut video timelines through undoable operations, verified assets, and frame-level visual validation.

38|4|Updated Jul 29, 2026
One-click install
npx skills add https://github.com/6174/recut --skill recut-editor-6174
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: recut-editor
Source: https://github.com/6174/recut/tree/main/service/skills/recut-editor
Command: npx skills add https://github.com/6174/recut --skill recut-editor-6174

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? It gives an AI agent a complete directing and editing contract for the Recut timeline editor, so new videos and revisions to existing timelines are executed through safe, undoable operations instead of ad-hoc project file edits. ## Core Features & Use Cases - Intent and route dispatch: Classifies each request as new authoring, timeline revision, visual/audio revision, asset replacement, or delivery, then routes to speech-led, media-led, generated-video, motion-graphics, or voice-led treatments. - Undoable timeline operations: All writes go through timeline.command with baseVersion conflict handling, project locks, checkpoints, and work.cancel rollback for interrupted sessions. - Three-layer verification: Requires structural proof via timeline.validate, settled-frame pixel proof via preview.frame/batch, and delivery proof via completed export jobs before claiming results. - Use Case: Ask the agent to turn a voiceover script into a narrated video: it builds a visual-first sync map, generates TTS audio, places motion graphics and B-roll on the timeline, validates structure and frames, then exports the final cut. ## Quick Start Ask the agent to create a short narrated video from your script and have it plan scenes, generate assets, place them on the timeline, and export the finished cut.

Frequently Asked Questions about recut-editor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I edit a video timeline with an AI agent in Recut?▼

The agent reads workflow.context and timeline.read to establish project state, determines intent and scope, then writes changes through timeline.command operations. Every edit is undoable via history.undo and validated with timeline.validate plus settled-frame previews before delivery.

How to add AI-generated video or motion graphics to a timeline?▼

Generated media uses recut.video.generate or recut.image.generate, returning a stable assetId that is placed on the timeline immediately. Motion graphics are created via recut.motion-graphic.create and must reach verified status before timeline.placeComponents inserts them.

Can the agent export videos without the editor open?▼

No. Preview and export require the editor iframe to be online; otherwise tools return editor-not-open. Headless mode is not implemented and returns headless-unavailable, so results can only be reported as drafts pending visual acceptance.

What happens when a timeline edit conflicts with another change?▼

Writes carry a baseVersion; on conflict the tool returns currentVersion and opsSince. The agent rereads the timeline, understands the external changes, and replays its own operations rather than overwriting the whole project.

How does the agent handle interrupted or cancelled editing sessions?▼

Each session starts with project.lock and work.checkpoint. On user correction, pending jobs are cancelled via recut.job.cancel and committed mutations are rolled back with work.cancel by checkpoint sequence, never by version-based undo.

Why must generated assets be verified before timeline placement?▼

Unverified mediaIds or componentIds are rejected by timeline.validate. Motion graphic jobs must complete with verified status, and all assetId, trackId, and elementId values must come from tool return values rather than being fabricated.