Cut as Code routes video requests to the right editing workflow and recovers project state without duplicating verified work.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install cut-as-code
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install cut-as-code using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Cut as Code Entry Point?
Cut as Code is the mandatory entry point for Project Protocol V1 video work. It determines the requested deliverable, checks the explicit project state when needed, and hands the task to the skill that owns the requested operation—such as cutting, color grading, captions, B-roll, or Shorts. This Openclaw Skills workflow keeps routing read-only and uses recorded project state rather than assumptions based on which files happen to exist.
For multi-operation requests, it coordinates prerequisites and a defined execution order, then directs the project to one shared main render after selected operations pass their checks. The entry point handles routing and recovery only; the selected leaf skill owns editorial decisions, review gates, rendering details, and self-checks.
Cut as Code Entry Point Use Cases
- Continue a named operation on an existing video project without repeating unrelated intake.
- Start from a source file or URL when the user has specified a clear deliverable.
- Route a request to cut, color grade, add B-roll, captions, content cards, motion graphics, or create Shorts.
- Coordinate multiple requested operations in the documented order and render the shared main delivery once.
- Compare a verified final delivery with the original or finalize approved short-form derivatives.
- Identify missing inputs, ambiguous deliverables, stale dependencies, or invalid project state before proceeding.
How Cut as Code Entry Point Works
- Identify whether the request names an operation on an existing project, supplies a project root, provides a source and clear deliverable, or lacks enough information to proceed.
- For an explicit project root, inspect
work/project.json,work/timeline.json, and only the selected operation's relevant plan or evidence. Treatwork/project.jsonas the shared manifest and determine progress from recorded status, revision, dependencies, checks, and render state. - Match the requested result—not incidental keywords—to its owning skill. If the deliverable is ambiguous or essential input is missing, ask one concise routing question.
- Read the selected skill's complete instructions and required sub-skills, then hand off. For multiple requested operations, freeze only those operations, run shared understanding once, and execute the selected work in its prescribed order.
- Re-read project state between handoffs. Do not run concurrent skills that could modify shared project or render artifacts, and do not repeat work that is already verified.
- After active operations pass their own gates, compile and render the shared main delivery once. Run comparison or Shorts finalization only against the verified delivery when requested.
Cut as Code Entry Point Setup
No standalone installation command or package manager is specified in the skill. Use Cut as Code as the entry point in an environment where this skill and its routed video skills are available.
- For an existing project, provide its explicit project root. The repository root is not automatically a video project.
- For new work, provide the source file or URL and the intended deliverable. The workflow routes initialization through
/video-understand. - Ensure the selected leaf skill and its required sub-skills are available; if one is missing, stop and report it rather than recreating its instructions.
- For rendering, the documented tools are
build_render_plan.pyandrender_project.pyfrom/video-understand. Run them only after selected active operations pass their gates.
# Inspect the shared manifest and timeline from the explicit project root
cat work/project.json
cat work/timeline.json
Cut as Code Entry Point Data Schema & Taxonomy
The workflow treats the explicit project root as the boundary of a video project and work/project.json as its only shared manifest. It reads work/timeline.json for timeline state and consults only the selected operation's plan or evidence when routing.
| Data or state | Purpose |
|---|---|
work/project.json |
Shared project manifest; records operation state and is re-read between handoffs. |
work/timeline.json |
Timeline state; changes can invalidate Shorts candidate approval. |
| Operation plans and evidence | Operation-specific planning and verification inputs; inspect only what the selected operation needs. |
| Operation metadata | Use status, revision, based_on, and check to establish progress and whether work is current and verified. |
| Render state | Determines whether a delivery exists and whether it is ready for comparison or derivative finalization. |
final.mp4 and final/shorts/ |
Shorts planning must not create these; approved derivatives are finalized and extracted only after the verified main delivery exists. |
Directory presence alone does not establish completion. Preserve existing user changes and artifacts, do not re-run verified work, and resolve missing or stale dependencies before resuming the requested operation.
Cut as Code Entry Point Advanced Features
- Routes by deliverable across understanding, cutting, color grading, B-roll, motion graphics, content cards, captions, comparison, and Shorts workflows.
- Supports project-state recovery based on manifest metadata, revisions, dependencies, checks, and render state.
- Coordinates multi-operation requests in canonical order: cut, color grade, B-roll, captions, content cards, then motion graphics.
- Uses a staged Shorts workflow: plan and review candidates before captions and overlays, then finalize approved derivatives after the main delivery is verified.
- Protects the visible face and head silhouette of primary or foreground people when placing content-card and motion-graphics overlays; cues must be repositioned, redesigned, or skipped if compliant placement is not possible.
- Allows valid all-skipped or zero-cue outcomes when permitted by the owning skill; it does not invent filler work.
- Keeps routing read-only and delegates editorial decisions, review gates, rendering details, and self-checks to the owning skill.
SKILL.md
Loading
Related Openclaw Skills
Email Design
Designs distinctive, on-brand emails and converts an approved creative brief into a preview-ready format for your sending platform.
Gauntlet
Gauntlet runs bounded, evidence-driven adversarial reviews to uncover substantiated defects and verify root-cause repairs.
Pet Care Planner
Pet Care Planner turns a cat or dog’s profile into a structured weekly feeding plan, health timeline, and concise risk reminders.
neo_ppt_skills
Generate presentation slides from a topic or article, then deliver both a polished image-based deck and an editable PPTX.
Chelaile Real-Time Bus
Query real-time bus arrivals, vehicle locations, routes, and public transit plans across 480 cities through Chelaile MCP.
Blave Quant
Blave Quant connects AI agents to global market data and exchange workflows, with explicit confirmation required for every trading action.
Featured*
