Skip to content
SkillModel-led2,901 tokens in use

gpt-image

An agent runbook for generating, editing and inpainting images with GPT Image 2, 2.5 and 2.6 through a packaged CLI, plus sprite-sheet GIFs and image-led websites.

6,166 stars on the repository MIT Licensewuyoscar/GPT-Image2-Skill · skills/gpt-image
claude plugin marketplace add wuyoscar/GPT-Image2-Skill && claude plugin install gpt-image@wuyoscar-skills --scope project
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing gpt-image costs

67tokens idle
2,901tokens in use
26,387tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works. Scripts: scripts/generate.py.

Touches

shelldetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure19/25
  • Examples12/15
  • Output15/15
  • Size13/15
  • Resources3/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Needs attention
  • Runs shell commands−5

    does this on the machine, and the frontmatter does not say so

    SKILL.md, line 48
  • Overrides the user's instructions−25

    tells the model to ignore its instructions or to hide what it does from the user

    references/craft.md, line 161
  • Overrides the user's instructions−25

    tells the model to ignore its instructions or to hide what it does from the user

    references/gallery-research-paper-figures.md, line 400
  • Hidden characters−20

    characters that render as nothing or reorder the text

    references/openai-cookbook.md, line 883
  • Hidden characters−20

    characters that render as nothing or reorder the text

    references/openai-cookbook.md, line 928

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity76
  • Maintenance72
  • Community44
  • Maturity28
  • Momentum40

This skill tells the agent how to produce images with OpenAI’s GPT Image models by calling a packaged command line tool, so no new API code gets written. It resolves which model to use (2.6 Flare by default), picks size and quality, and routes the call to text-to-image, reference edit or masked inpaint.

It also covers native GIFs, turning an evenly divided sprite sheet into an animated GIF locally, and building a real HTML page around generated assets. A reference gallery of prompts sits beside it.

What You Get

  • Images written to disk from a text prompt, one or many reference images, or a prompt plus an alpha mask
  • A model table covering GPT Image 2, 2.5 and 2.6 Flare and Sunburst variants with their API IDs
  • Native GIF output, and a local sprite-sheet to GIF step that needs no API call
  • A gallery of prompts with reference images, grouped by category, to borrow from

Common Use Cases

  • Make a poster or typography-heavy graphic at a chosen size and quality
  • Edit a photo with reference images or inpaint a masked region
  • Generate a sprite sheet and slice it into a looping GIF with a preview player
  • Create visual assets and build an interactive web page around them

What Makes It Unique

The skill drives one CLI with an explicit model flag, keeps cost discussion apart from the image prompt, never silently switches models or rewrites the prompt after a failure, and loads only one small reference slice per task. It adds animation and website workflows that most image skills leave out.

Under The Hood

The skill is mainly instructions plus a small Python launcher that finds the gpt-image command, or falls back to uv or uvx to fetch it from GitHub. The CLI sends your prompt and any reference images or masks to the OpenAI images API using your API key, which it reads from the environment, a local env file or a home-directory env file, and writes the returned images to disk. Calls can be billed to your OpenAI account. The agent is told not to write keys or reinstall anything without being asked. The automated scan also flags a few reference files: some contain example prompt text that reads like an instruction to ignore earlier rules, and one contains invisible characters. These sit in the prompt gallery and cookbook material, which the agent loads only when it needs an example.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers