codex-pet

Generate Codex-compatible pet spritesheets and manifests from a single reference image.

12|2|Updated Aug 12, 2026
One-click install
npx skills add https://github.com/genmedia-labs/skills --skill codex-pet-genmedia-labs
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: codex-pet
Source: https://github.com/genmedia-labs/skills/tree/main/codex-pet
Command: npx skills add https://github.com/genmedia-labs/skills --skill codex-pet-genmedia-labs

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires @runcomfy/cli, imagemagick.

What problem does it solve? Creating a custom Codex Pet normally requires Codex Pro and the internal $imagegen system skill. This Skill produces the exact same pet.json manifest and 1536x1872 spritesheet.webp atlas using only a RunComfy token, the RunComfy CLI, and ImageMagick, so anyone can install a custom desktop pet for OpenAI Codex. ## Core Features & Use Cases - Single-call generation: Calls OpenAI GPT Image 2 edit once via the RunComfy CLI to produce a canonical chibi pose on a magenta chroma-key background. - Deterministic atlas assembly: Uses ImageMagick micro-transforms (shifts, blinks, rotations, mirroring) to build all 9 animation rows (idle, running, waving, jumping, failed, waiting, review) into the 1536x1872 atlas Codex expects. - Automatic installation: Writes pet.json and spritesheet.webp directly into ${CODEX_HOME:-$HOME/.codex}/pets/<name>/ where Codex picks them up next to the 8 built-in pets. - Use Case: Provide a photo of your cat, and within about two minutes have a pixel-art chibi cat pet installed in Codex that animates while Codex codes. ## Quick Start Ask the agent to create a Codex pet from your source image URL and install it into your Codex pets directory.

Frequently Asked Questions about codex-pet

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a custom Codex Pet without Codex Pro?▼

Use this Skill with the RunComfy CLI and a RUNCOMFY_TOKEN. It calls GPT Image 2 edit once to make a canonical chibi pose, then assembles the full spritesheet with ImageMagick, requiring no Codex Pro or $imagegen access.

What is the difference between this skill and OpenAI's hatch-pet?▼

Both produce the identical pet.json and 1536x1872 spritesheet.webp artifact. Official hatch-pet uses the Codex-internal $imagegen skill requiring Codex Pro, while this skill runs through the RunComfy CLI with only a RUNCOMFY_TOKEN.

What image formats and sizes does a Codex Pet spritesheet require?▼

The spritesheet must be a 1536x1872 PNG or WebP with 8 columns and 9 rows of 192x208 cells on a transparent background. Each row maps to one animation state with a fixed number of leading frames.

Can I use a pet, object, or non-human subject as the source image?▼

Yes, pets, mascots, objects, and foods all work. The GPT Image 2 prompt simplifies any source into the chibi Codex Pet house style automatically, though small identity details may drift.

Why does my Codex Pet lose body parts after chroma-keying?▼

The magenta #FF00FF chroma-key removes any near-magenta colors in the pet itself. Increase the fuzz tolerance carefully or switch the chroma-key color to cyan or green in both the prompt and post-processing.

What are the limitations of the ImageMagick animation approach?▼

Animation uses micro-transforms of one canonical pose, so motion is subtle (1-2 px shifts, blinks, tilts) rather than dramatic frame-by-frame movement. This matches Codex's built-in pets but cannot produce distinct running cycles.