---
name: animate-rigged-art
description: Turn a supplied flat image or a visual concept into a layered, editable 2D skeletal animation with DragonBones/LoongBones data, an HTML preview, visual quality gates, revision handling, and project export. Use for requests to rig, puppet, layer, animate, loop, or export 2D characters and articulated artwork.
---

# Animate Rigged Art

Create one approved animation at a time. Keep the original appearance as the visual source of truth.

## Workflow

1. Inspect the supplied image and motion request. If no image exists, plan a rig-ready master image using the host's image-generation capability.
2. Grill only decisions that can change the artifact: subject, moving parts, motion, duration, loop behavior, camera, canvas, background, style invariants, reference motion, and export needs. Ask one focused question at a time and recommend a default.
3. Write `brief.json`, show the compact brief, and get explicit approval before generating layers or a rig.
4. If generating artwork, create and obtain approval for the master image first. Prefer a clear base pose, visible joints, minimal self-occlusion, and a fixed camera.
5. Read [project-contract.md](references/project-contract.md). Create isolated RGBA layers, reconstruct pixels hidden behind overlaps, and define pivots, bones, z-order, and animation keys in `project.json`.
6. Recompose the neutral pose and compare it with the approved source. Do not continue while identity, silhouette, colors, or proportions visibly drift.
7. Read [quality-gates.md](references/quality-gates.md). Run a builder/critic loop for asset fidelity, rig deformation, motion, and export. Give each fresh critic the approved brief, reference, and actual rendered artifact—not the builder's explanation. Let only the builder modify files.
8. Run the bundled script with the active skill directory as `<skill-dir>`:

   ```text
   python <skill-dir>/scripts/rigforge.py build <project-dir>
   python <skill-dir>/scripts/rigforge.py validate <project-dir>
   ```

9. Serve `<project-dir>/dist` over HTTP and show the user the live preview. Convert each comment into a scoped change and resume only the earliest affected stage.
10. After approval, export the editable source project and requested rendered formats. Package the source project with:

   ```text
   python <skill-dir>/scripts/rigforge.py package <project-dir>
   ```

## Boundaries

- For v1, accept one articulated 2D subject, a fixed camera, and motion close enough to the base view that a cutout rig can represent it.
- Do not force skeletal rigging onto logos, fluids, particle effects, major camera turns, or photorealistic talking heads. Explain the mismatch and propose the appropriate animation method.
- Never invent permission to animate a real person's likeness. Confirm consent when identity or impersonation risk is material.
- Preserve the original and approved versions. Do not overwrite user source files.
- If image generation/editing or fresh critic contexts are unavailable, state the reduced capability and continue only with inputs the host can safely process.