# Generate Pose-Controlled Images and Videos

Pose and motion references are difficult to route across image and video generation models. This skill selects a suitable RunComfy model and prepares the generation request.

## Install

```bash
npx skillstore add agentspace-so/controlnet-pose
```

## Metadata

- Status: approved
- Slug: agentspace-so-controlnet-pose
- Skillstore revision: r2
- Version status: missing
- Tree hash: e56782ba7e2727ff222b00fa5178725b32deb9cffd3b729277d48c74488599f7
- Author: agentspace-so
- GitHub username: agentspace-so
- License: MIT
- Repository: https://github.com/agentspace-so/runcomfy-agent-skills/tree/main/controlnet-pose/
- Ref: 181fdefcafd96b041926e61c4b2e306ca7e7820e
- Supported tools: Claude, Codex, Claude Code
- Audit status: complete
- Agent install advisory: confirmation\_required
- Manual install advisory: allowed
- Artifact signature: available
- Audit attestation: unavailable
- Human verification: not\_verified
- Risk factors: external\_commands, network, filesystem
- Quality score: 38
- Quality tier: warning
- Public page: https://skillstore.pages.dev/skills/agentspace-so-controlnet-pose
- Manifest: https://skillstore.pages.dev/api/skills/agentspace-so-controlnet-pose/manifest

## Capabilities

- Classifies requests as video motion transfer or still-image pose conditioning.
- Selects Kling Motion Control Pro or Standard for reference-video motion transfer.
- Selects Wan 2-2 Animate for stylized character animation.
- Selects Z-Image Turbo ControlNet LoRA for pose, depth, or canny image control.
- Prepares RunComfy CLI requests with model inputs and output directories.
- Directs complex multi-condition requests to hosted ComfyUI workflows.

## Use Cases

- Transfer Performance Motion: Apply a reference performance video to a target character with Kling Motion Control.
- Lock an Image Pose: Generate a new image from an OpenPose, DWPose, depth, or canny control image.
- Prototype Character Animation: Route stylized animation drafts to Wan or lower-cost Kling models before final rendering.

## Prompt Templates

### Create a Pose-Controlled Image

```
Create an image of [subject] in [setting] using this [OpenPose or DWPose] control image: [URL]. Save the result in [folder].
```

### Transfer Motion to a Character

```
Transfer the motion from [reference video URL] to [character image URL]. Use [Pro or Standard] quality and save the generated video in [folder].
```

### Choose a Model for a Draft

```
Compare suitable RunComfy routes for [project]. The input is [media type], the style is [style], and the control reference is [type and URL].
```

### Plan a Multi-Condition Workflow

```
Plan a workflow using [pose], [depth], and [reference image] controls for [goal]. Identify whether CLI routes suffice or a ComfyUI workflow is required.
```

## Limitations

- Requires the RunComfy CLI, an authenticated account, and network access.
- Reference videos, character images, and control images must be available through usable URLs.
- Direct CLI routes support single-condition workflows, not complex ControlNet stacks.
- Generation quality depends on reference framing, model behavior, and third-party service availability.

## Best Practices

- Use clean, well-framed references with one clear subject and continuous motion.
- Confirm the control type matches the selected model or LoRA.
- Review service privacy requirements before sending prompts or media URLs.

## Anti Patterns

- Do not use still-image routes for video motion transfer.
- Do not send multi-condition stacks to a single-condition CLI endpoint.
- Do not interpolate untrusted text directly into a shell command.

## Security Audit

- Audited at: 2026-07-23T08:39:20.112\+00:00
- Summary: The static analyzer mostly mistook Markdown syntax, documentation links, and placeholder URLs for executable behavior. Two contextual risks remain: unpinned package execution and shell examples that can become injectable when an agent substitutes unescaped user input.

## Stats

- Views: 49
- Downloads: 7
- Favorites: 0
- Popularity score: 0