GPT Image 2by ConardLi

GPT Image 2 Skill.

An image workflow that distinguishes having a prompt from having a working renderer.

Image Generation
Prompt Design
Image Editing

Source Review · GPT Image 2SKILL.md

Task Image Generation Produces • A task-specific image prompt with explicit subject, composition, and text requirements • A generated or edited image when an available backend completes the request • Saved prompts and a record of whether the API, native tool or prompt-only mode was used Before You Start Local API mode requires explicit Garden enablement, Node scripts, and configured credentials. Native mode depends on the host image tool. Model access, pricing, and supported parameters must be checked in that environment.

Editorial summary · source reviewed 2026-10-07 · not a generated example

01 What It Does

GPT Image 2

A template-led image skill with separate local API, host-native, and prompt-advisor paths, plus explicit file and mode reporting.

Source-based workflowNot an execution benchmark

The Working Approach

This GPT Image 2 skill belongs to ConardLi's Garden collection. It combines task-specific prompt templates with a capability check that decides how an image can actually be produced. The templates cover visual tasks such as product imagery, interface mockups, posters, diagrams, and illustration. Their structured fields describe the intended image; they are not API request bodies to paste unchanged into a provider endpoint.

The workflow has three execution paths. A deliberately enabled local API path uses the repository's scripts and credentials. A host-native path hands the prepared prompt to an image tool already available in the agent environment. If neither exists, the skill acts as a prompt advisor and must not say it generated an image. That distinction matters when the same instructions are used across different installations.

Expected Deliverables

A task-specific image prompt with explicit subject, composition, and text requirements
A generated or edited image when an available backend completes the request
Saved prompts and a record of whether the API, native tool or prompt-only mode was used

See the Example Brief

From Brief to Review

01
Identify The Available Mode
Check whether the Garden API path is explicitly enabled, whether a native image tool exists, or whether the environment can only prepare a prompt.
02
Choose A Relevant Template
Load the prompt structure that fits the requested visual. Ask for concrete missing inputs such as product details or exact visible text instead of collecting unrelated style preferences.
03
Render Or Advise
Use the configured backend, supply the prompt to the host tool, or deliver a clearly labeled prompt-only result. Editing also needs the original image and any intended mask.
04
Record The Actual Result
Keep the prompt alongside available image output. Report which files exist and avoid presenting an unavailable rendering mode as a successful generation.

02 When to Use It

When to Use It

Local API mode requires explicit Garden enablement, Node scripts, and configured credentials. Native mode depends on the host image tool. Model access, pricing, and supported parameters must be checked in that environment.

A Good Fit

Structured Visual BriefsUseful when an image needs exact labels, a deliberate composition, or product constraints that a short decorative prompt would omit.
Mixed Agent EnvironmentsThe explicit fallback helps a team understand why one installation can render while another can only prepare the same request.

Limits to Check

No Backend IncludedThe skill does not provide API credentials, provider access, or a guarantee that a named model is available in the current environment.
Raster OutputA generated diagram or interface mockup is an image, not editable vector geometry or working UI code. Choose a different production path when those artifacts are required.

Bring These Inputs

Visual TaskName the output type, intended use, aspect ratio, and subject. Include exact words that must appear inside the image.
Reference AssetsSupply product references or the original image for an edit, and state what may change versus what must remain recognizable.
Execution ChoiceSpecify whether to use a configured API, an available native image tool, or prompt preparation only. Keep credentials outside the prompt.

03 Decisions That Matter

Decisions That Matter

Report the Actual Output
A prompt-only result can still be useful, but it is a different deliverable from an image. This workflow makes that limitation visible before users search for a nonexistent file.
Templates Are Briefs
Separating composition fields from transport details allows a structured visual idea to be adapted to a native tool without pretending every backend accepts identical parameters.

04 Try a Focused Brief

Try a Focused Brief

The example brief below has not been tested. Check the skill’s setup before using it.

Review the Setup

Install or link the Garden repository skill with its templates and scripts. The reviewed source does not establish a universal installation command.

Upstream Source File

https://github.com/ConardLi/garden-skills/blob/aaf9a82f5efd73e87cc0998edc398e75bfc35901/skills/gpt-image-2/SKILL.md

Start With One Deliverable

Adapt the scope, source material and constraints in this brief before running the skill.

Example Brief

Use the Garden GPT Image 2 skill to prepare a product-comparison illustration for the supplied two desk lamps. Keep both shapes recognizable, use a neutral studio background, and include no text or invented logos. Check the available execution mode first. If a native image tool is available, use it; otherwise deliver the complete prompt and clearly say that no image was rendered.

Inspect the Result

Check the actual output against these observable criteria.

The delivery states whether an image was rendered or only a prompt was prepared.
Any reported image file exists and reflects the requested subject and composition.
The saved prompt contains the exact constraints without exposed API credentials.

OpenDesign provides a local design workspace for coding agents. This listing does not mean the skill is bundled or that its external tools, credentials and software dependencies are included.

Download OpenDesign Read the Quickstart

05 Source and License

Source and License

This page describes the pinned source revision linked below. Upstream behavior and requirements can change.

Source Owner: ConardLi
Repository: ConardLi/garden-skills
Skill File: skills/gpt-image-2/SKILL.md
License: MIT
Reviewed Snapshot: 2026-10-07 · aaf9a82f5efd

Editorial Source Summary

A template-led image skill with separate local API, host-native, and prompt-advisor paths, plus explicit file and mode reporting.

Original summary by OpenDesign, based on the linked source. Repository stars are a dated popularity signal, not skill usage or a quality score.

06 Questions Before You Start

Questions Before You Start

Does every mode need an API key?

No. The local Garden API mode requires its configured credentials and enablement flag. A native tool uses its host's setup; prompt-only mode does not render.

Are the JSON templates ready-made API bodies?

No. They are prompt structures. The chosen execution path handles the actual request format.

Can it deliver an editable technical diagram?

Its image path produces raster artwork. The technical-diagram template should not be mistaken for editable SVG, a diagram source file, or a CAD model.

OpenDesign

Make Room for the Next Design Iteration.

Keep your brief, source material and design work together in a local workspace. Check the skill’s own setup before adding it to your workflow.

Download OpenDesign Browse Design Skills