gpt-image-2-content-factory

name	gpt-image-2-content-factory
description	GPT Image 2 content factory and prompt/API workflow skill. Use when the user asks for OpenAI gpt-image-2 prompts, commercial visual content, ecommerce detail pages, ads, posters, character sheets, landing page visuals, logos, app UI mockups, brand visual systems, icon systems, reference-image edits, batch variants, prompt diagnosis, API examples, or parameter choices for gpt-image-2 only.

GPT Image 2 Content Factory

Use this skill only for OpenAI gpt-image-2.

This merged skill combines:

• gpt-image-2-auto-generator: prompt/API boundaries, parameter validation, reference-image and mask edit workflows.
• gpt-image-2-content-factory: seven-engine content pipeline, nine commercial visual scenes, delivery cards, iteration QA.

Do not route, compare, mention, or generate variants for other image models. If the user asks for another model, state that this skill is scoped to gpt-image-2 and provide only a gpt-image-2 version if useful.

Source Of Truth

Use current OpenAI documentation for version-sensitive API details:

• https://developers.openai.com/api/docs/models/gpt-image-2
• https://developers.openai.com/api/docs/guides/image-generation
• https://developers.openai.com/api/reference/resources/images

Use local files as workflow guidance:

• API and platform boundaries: shared/gpt-image-2-platform-spec.md
• Engineering examples: references/gpt-image-2-engineering-spec.md
• Prompting: references/gpt-image-2-playbook.md
• Content pipeline: engines/e01-brief-decoder.md through engines/e07-iteration-qa.md
• Nine commercial scenes: shared/nine-scenes-codex.md
• Prompt recipes: shared/prompt-recipes.md
• Anti-patterns: shared/anti-patterns.md
• Visual vocabulary: references/visual-vocabulary.md, references/visual-variables-core.md
• Examples: examples/

Operating Modes

Mode	Trigger	Output
M1 FULL	User gives a product, brand, idea, or rough brief and wants a full visual plan	Brief decode, scene route, visual strategy, prompt, parameters, API code, iteration plan
M2 SCENE	User names one scene such as ecommerce page, poster, logo, UI, or icon system	Single-scene prompt package
M3 EDIT	User gives reference image(s), mask, or edit instruction	Reference analysis, edit prompt, Image API edit parameters
M4 BATCH	User asks for multiple variants, SKU set, angles, or batch output	Batch prompt matrix and n/variant strategy
M5 DIAGNOSE	User gives an existing prompt or bad output and asks to improve it	IQS score, defects, rewrite, and concrete diff
M6 SERIES	User asks for a family of assets under one visual DNA	Cross-scene prompt matrix with shared brand anchors

Nine Scenes

ID	Scene
S1	Ecommerce detail page
S2	Ecommerce ad or campaign banner
S3	Recommendation poster
S4	Character sheet
S5	Landing page visual
S6	Logo exploration or logo grid
S7	App UI mockup
S8	Brand visual identity system
S9	Icon system

Execution Flow

1. Decode the brief with E01. If object, scene, or ratio is missing, ask at most two focused questions; infer low-risk defaults from context.
2. Route mode and scene with E02.
3. Build the visual strategy with E03.
4. Forge a Chinese prompt and an English prompt with E04, keeping both versions semantically identical.
5. Compile valid gpt-image-2 parameters with E05.
6. Deliver with E06.
7. Run E07 plus shared/anti-patterns.md before final output.

API Boundaries

• Always use model: "gpt-image-2" for Image API calls.
• Use v1/images/generations for text-to-image.
• Use v1/images/edits for reference images, edits, and mask-guided edits.
• Parse data[0].b64_json; do not rely on URL returns.
• Use quality: "low" | "medium" | "high" | "auto".
• Use output_format: "png" | "jpeg" | "webp".
• Use background: "auto" or opaque output. Do not use background: "transparent" for gpt-image-2.
• Do not use input_fidelity; gpt-image-2 handles image inputs at high fidelity automatically.
• Validate sizes with the four constraints in shared/gpt-image-2-platform-spec.md.
• For masks, require identical dimensions and format between source image and mask, file size under 50 MB, and an alpha channel in the mask.

Delivery Standard

Default output:

## GPT Image 2 Visual Package

Scene:
Mode:
Assumptions:

Parameters:
- model: gpt-image-2
- size:
- quality:
- output_format:
- background:
- n:
- references/mask:

Chinese Prompt:
...

English Prompt:
...

API Example:
...

QA Notes:
- ...

Iteration Options:
- ...

Use code examples only when the user asks for API use, engineering integration, or a runnable package.

Final Check

Before final delivery:

• Every model mention is gpt-image-2 or gpt-image-2-2026-04-21.
• No third-party model, older image model, or text-model orchestration guidance remains in the answer.
• Parameters are valid for gpt-image-2.
• Size satisfies the four constraints.
• Text to render is quoted exactly.
• Reference-image instructions say what to preserve and what may change.
• Mask edits include mask requirements.
• No pricing, rate limit, availability, or policy statement is made unless verified from current OpenAI docs.