Pixeltamer GPT Image Skill — AI skill for Claude Code

pixeltamer — image generation skill for AI coding agents | gpt-image-2 toolkit

Six different image-gen skills were rotting in my `to_check/` folder, each doing one thing well and three things badly. So I merged them, kept the parts that earned their keep, dropped the rest. This is the result — one skill, two backends, three modes, zero generic-AI-slop output.

Generate, edit, and compose images with `gpt-image-2` from inside Claude Code, Codex CLI, OpenCode, Cursor — or any agent that reads `SKILL.md`. Bring an OpenAI API key OR sign in to your ChatGPT subscription via codex; same prompts, same recipes, your call on how to pay. Ships with a prompting doctrine that drops the "8K, ultra detailed, masterpiece" magic words this model actively hates, seven production recipes (infographic, meta-ad, viral-linkedin, ui-mockup, editorial-cover, product-photo, mascot — the new mascot recipe has a phased Discovery → Concept Lock → Production → Maintenance workflow with empirically-validated format choices), a curated `playbook/` of remixable prompts for character-sheets and typography-posters, a central `references/index.md` routing map so the right recipe gets loaded on first try, and a 39-test suite covering the multi-image batch verifier plus the index-staleness check. Zero runtime dependencies — every backend and helper runs on stock bash / Python stdlib / Node, nothing to `npm install`.

Works in Claude Code, Codex CLI, OpenCode, Cursor, and 50+ other agents via [the Skills CLI](https://github.com/vercel-labs/skills).

---

Install

Quick install (recommended)

npx skills add gabelul/pixeltamer-gpt-image-skill

One command. Auto-detects your agent, symlinks the skill into the right place, works across all of them at once. Update later with `npx skills update`.

Manual setup