layered-img-skill

KO | EN

gitlite — search

#javascript #python #ai #llm #android #deep-learning #linux #docker #macos #java #cli #vue #awesome-list #openai

layered-img-skill

(★ 9)

a Codex skill for turning a complex source image into a set of demo-friendly, editable image layers.

layered-img-skill Latest Version Download

Download Latest Version (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# layered-img Skill

`layered-img` is a Codex skill for turning a complex source image into a set of
demo-friendly, editable image layers. It uses a plan-first workflow, generates
standalone assets on a pure green sprite sheet, crops those generated assets into
transparent PNGs, and writes bbox/z-index metadata for assembly.

This is not pixel-perfect PSD recovery. The goal is to produce convincing,
separately draggable image layers that are practical for demos, editing, and
presentation exports.

## What It Produces

A normal run creates a work folder like this:

```txt
layered-img-work/<run-name>/
  input/source.png
  prompts/layer-plan.json
  prompts/batch-001-prompt.txt
  generated/batch-001-green.png
  layers/batch-001-crops/
  layers/assets/
  assembly/layer-compose.json
  manifests/
  scripts/generate_assembly_ppt.mjs
  preview/
  output/layered-composition.pptx
```

The important deliverables are:

- Transparent PNG layer assets in `layers/assets/`.
- `assembly/layer-compose.json` with canvas size, layer paths, source-image bbox
  placement, and z-index.
- An optional PowerPoint export at `output/layered-composition.pptx`, where each
  layer is a separate draggable image object.

## Install

From this repository:

```bash
./install.sh
```

By default, the installer copies the skill into:

```txt
~/.agents/skills/layered-img
```

To install somewhere else:

```bash
AGENTS_SKILLS_DIR=/path/to/skills ./install.sh
```

Restart Codex or reload skills if the skill does not appear immediately.

## Usage

Ask Codex to use the skill on an image, for example:

```txt
Use layered-img to split /absolute/path/to/source.png into editable layers.
```

The skill workflow is encoded in `SKILL.md`. In short, Codex will:

1. Create a structured run folder with `scripts/init_layered_workdir.py`.
2. Write `prompts/layer-plan.json` before generation.
3. Write a batch prompt for a pure `#00ff00` green sprite sheet.
4. Generate `generated/batch-XXX-green.png` with image generation.
5. Crop the generated green sheet into transparent PNG assets.
6. Build `assembly/layer-compose.json`.
7. Optionally create a DeckKit-based PPTX assembly export.

## Manual Script Flow

Initialize a work folder:

```bash
python3 scripts/init_layered_workdir.py /absolute/path/to/source.png --name demo
```

After creating `prompts/layer-plan.json` and generating a green sprite sheet,
crop the generated sheet:

```bash
python3 scripts/crop_pure_green_assets.py \
  layered-img-work/<run-name>/generated/batch-001-green.png \
  --out-dir layered-img-work/<run-name>/layers/batch-001-crops \
  --tolerance 3 \
  --min-area 800 \
  --padding 12 \
  --bridge 0 \
  --prefix batch_001_asset
```

Copy or save the crop manifest to:

```txt
layered-img-work/<run-name>/manifests/crop-manifest-batch-001.json
```

Then build the normalized layer manifest:

```bash
python3 scripts/build_layer_manifest.py layered-img-work/<run-name>
```

To create the optional PPTX assembly script:

```bash
python3 scripts/write_deckkit_script.py layered-img-work/<run-name>
```

Then install and run from inside the work folder:

```bash
cd layered-img-work/<run-name>
npm install
npm run generate
```

## Important Boundary

`generated/batch-XXX-green.png` must come from image generation. Do not create it
with local source-image crops, segmentation masks, alpha masks, OpenCV/PIL
cutouts, or manual compositing.

The bundled scripts may initialize folders, crop an already-generated green
sheet, build manifests, and assemble/export layers. They must not manufacture the
green sprite sheet from the original image.

## Repository Layout

```txt
.
  SKILL.md                         Skill instructions used by Codex
  install.sh                       Local skill installer
  agents/openai.yaml               Agent marketplace metadata
  scripts/init_layered_workdir.py  Run-folder initializer
  scripts/crop_pure_green_assets.py
  scripts/build_layer_manifest.py
  scripts/write_deckkit_script.py
```

## Dependencies

Python scripts require:

- Python 3.10+
- Pillow
- NumPy

The optional PPTX export requires Node.js and installs:

- `@artifact-kit/deckkit`
- `@artifact-kit/deckkit-pro`

## License

Add a license before publishing broadly if this repository is intended for public
reuse.

layered-img-skill

layered-img-skill Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

layered-img-skill

layered-img-skill Latest Version Download

Today's picks

Was this content helpful?

Recent Feedback

Download README