About this skill

The Problem

Image generation calls often drift when a user asks for 16:9, 3 images, and stable output: the skill may execute the wrong number of calls, convert aspect ratios inconsistently, or repeat requests during retries. zimage turns text-to-image generation into a controlled pipeline: it fills in the required prompt, maps aspect ratios to explicit size values, confirms the planned call count, and then invokes the fixed endpoint without extra calls.

How It Works

Key behaviors include:
- Input normalization: prompt is required; size accepts width*height, defaulting to 1024*1536
- Ratio mapping: for example, 16:9 maps to 1536*864, and 9:16 maps to 864*1536
- Count control: one call by default; if the user requests N images, it executes exactly N times
- Key reuse: reads from the current conversation, then X_API_KEY, then local persisted config
- Structured output: returns parameter summary, planned vs actual calls, API result summary, and next-step guidance on failure

Boundaries

This skill is for text-to-image requests such as posters, covers, illustrations, and images with visible text. It does not handle image-to-image, inpainting, object removal, outpainting, OCR, video, or audio work. Parameter validation must complete before the API call, x-api-key should be masked in logs, and errors such as 401, 429, or 5xx should trigger targeted user guidance.

Screenshots

Use Cases

Product operations preparing marketplace assets can generate four `9:16` e-commerce images with Chinese titles using a fixed prompt and count.
Content editors creating official account covers can ask for two `16:9` cover images and keep the prompt and output count consistent.
Designers needing poster drafts can provide a Chinese prompt and let the skill default to a `1024*1536` vertical image size.
Product engineers validating an image skill can confirm that it runs the requested number of calls and returns parameter and result summaries.

Best For

Operations owners responsible for product and campaign materials need fixed-ratio image outputs that include Chinese or English visible text accurately.
Content editors producing web or social covers need a specified number of `16:9` outputs with stable prompts and controlled call counts.
Product engineers validating image generation endpoints need clear parameter summaries, executed call counts, and actionable failure guidance from each run.
Full-stack engineers maintaining agent workflows need reusable `X_API_KEY` handling and fewer manual repetitions before each image request across environments.