# Edit Video Regions with RunComfy

Unwanted objects and inconsistent regions can distract viewers across an entire video. This skill guides RunComfy model selection and regional editing prompts while identifying tasks that require separate masking workflows.

## Install

```bash
npx skillstore add prime-skills/video-inpainting
```

## Metadata

- Status: approved
- Slug: prime-skills-video-inpainting
- Skillstore revision: r1
- Version status: missing
- Tree hash: 67e4e7cd0854db094b314c7dcf7bccf19f270d35871979de5ad98769b6383b76
- Author: prime-skills
- GitHub username: prime-skills
- License: MIT
- Repository: https://github.com/prime-skills/runcomfy-agent-skills/tree/fca19ae084c2ae4667f5c8affbb99da5103d5832/video-inpainting
- Ref: a53bacaa2e85e44d9f929fa7a6d863cbe7fd1c03
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: allowed
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 70
- Public page: https://skillstore.pages.dev/skills/prime-skills-video-inpainting
- Manifest: https://skillstore.pages.dev/api/skills/prime-skills-video-inpainting/manifest

## Capabilities

- Recommends Wan, Lucy, or Seedream routes based on regional edit intent, identity preservation, and frame-sequence requirements.
- Provides a RunComfy CLI example for submitting a source video URL and an edit prompt to Wan.
- Guides spatial targeting, preservation instructions, and separate passes for individual edits.
- Identifies ComfyUI workflows for pixel-precise masks and temporal propagation outside the CLI.
- Explains CLI authentication, output-directory selection, and common exit codes.

## Use Cases

- Clean Up Background Distractions: Plan a regional edit to remove overhead cables or a passing background object while preserving the surrounding scene.
- Plan Consistent Product Replacements: Compare prompt-driven replacement, identity-aware restyling, and separate frame workflows for an authorized product video update.
- Select a Masking Workflow: Determine when a scientific or training clip needs precise masks and temporal propagation instead of a prompt-only model call.

## Prompt Templates

### Choose a Basic Cleanup Route

```
I own this clip: [video URL]. Recommend a route to remove the overhead cable while preserving everything else. Do not submit a job yet.
```

### Prepare a Regional Edit

```
Prepare a Wan edit for [video URL] to remove the box in the lower-left corner. Preserve surrounding content. Show the prompt and execution plan before submission.
```

### Compare Identity-Sensitive Routes

```
Compare Wan and Lucy for replacing the jacket in [video URL] while preserving the subject's identity. Identify limitations and recommend a small test before submission.
```

### Design a Precise Multi-Pass Workflow

```
Plan separate edits for [video URL]: remove a moving object, then replace a sign. Compare prompt-only editing with tracked ComfyUI masks. Include temporal checks and shell-safe input handling.
```

## Limitations

- Requires an external RunComfy CLI installation, authentication, network access, and access to the selected service model.
- Prompt-driven edits do not guarantee pixel-precise targeting or unchanged surrounding content.
- Mask propagation requires separate GUI workflows; frame extraction and video re-encoding require separate tools.
- This package contains guidance only; external CLI security claims and model performance were not verified.

## Best Practices

- Use source URLs explicitly provided for the task, and confirm authorization before sharing footage with an external service.
- Describe one spatially specific change per pass, request preservation, and inspect motion continuity after each edit.
- Serialize inputs and use argument-array execution where available; do not interpolate prompts or URLs directly into shell commands.

## Anti Patterns

- Combining unrelated regional edits into one prompt and expecting unchanged surrounding frames.
- Treating prompt-driven endpoints as substitutes for precise masks or guaranteed temporal propagation.
- Assuming JSON prevents shell injection or treating unverified CLI privacy claims as established guarantees.

## Security Audit

- Audited at: 2026-11-01T14:41:03.372\+00:00
- Summary: All 57 static findings are false positives involving Markdown, reference URLs, standard authentication documentation, or model-selection advice. One semantic concern remains: the guide incorrectly treats JSON input as protection against shell injection during command construction. No evidence found of malicious instructions, credential exfiltration, or security-review manipulation; external CLI behavior was not verified.

## Stats

- Views: 0
- Downloads: 0
- Favorites: 0
- Popularity score: 0