Install or update GPT Image Skill

For the coding agent carrying out the user's request. Use this guide for both a first install and an update. Run the commands, inspect the results, and report what is available. Keep the user's existing installation scope and API settings.

1. Identify the existing installation

Check the current runtime's plugin/skill list, command -v gpt-image, and, if uv is available, uv tool list. Inspect only the relevant skill folders and installer records. Do not print environment variables, API keys, or credential files.

• Claude Code plugin gpt-image@wuyoscar-skills: use the Claude update below.
• Skills managed by skills: keep their project or global scope.
• Codex native plugin: use the native plugin instructions in install and update.
• A standalone CLI managed by uv: update it separately if present.
• Hand-copied or locally edited files: preserve them. Use their original installer or a clean linked checkout; do not overwrite them to force an update.

2. Update an existing installation

Run only the commands matching the installations found above:

# Claude Code plugin
claude plugin update gpt-image@wuyoscar-skills

# Project-local skills installed by skills
npx --yes skills@latest update gpt-image get-prompt-from-image --project

# Standalone CLI installed from this repository by uv
uv tool upgrade gpt-image-cli

For skills originally installed globally, replace --project with --global. Do not update unrelated skills. A plugin normally includes the CLI source; a copied skill can delegate to an independently installed CLI, which may also need updating. Use the installation guide for other sources.

3. First install

Use the user's current runtime. For Claude Code, install the plugin at project scope unless the user requested a global installation:

claude plugin marketplace add wuyoscar/GPT-Image2-Skill
claude plugin install gpt-image@wuyoscar-skills --scope project

For Codex, install the two skills into the current project:

npx --yes skills@latest add wuyoscar/GPT-Image2-Skill --skill gpt-image get-prompt-from-image --agent codex --copy

The skills installer requires Node.js 22.20 or later. Other supported runtimes can replace codex with their agent name. If the user only wants the CLI, or the skill needs a backend, use an existing uv installation to run it without a global install:

uvx --refresh --from git+https://github.com/wuyoscar/GPT-Image2-Skill gpt-image --help

If a required package manager is absent, explain the missing dependency and use the user's approved setup method. Do not silently switch to a global install.

4. Verify and report

Confirm that the intended plugin or skill is listed in the selected scope. For an installed CLI, run gpt-image --version, gpt-image --help, and gpt-image animate --help. For a plugin's bundled CLI, run the same flags through uv run <plugin-root>/skills/gpt-image/scripts/generate.py; package metadata may be absent there, so also inspect the plugin manifest version.

Report the installed source, scope, version and checks. Tell the user to start a new agent session to load updated instructions. Installation does not require an API key or a paid image request. Existing API-key settings must remain unchanged; if the user later requests generation, check availability without revealing the key. Never claim a successful image call based only on installation or --help.

The single-guide entry point was inspired by Agent Reach's install/update guide.