cover-image

---
name: cover-image
description: Generate a cover image by running Gemini and DALL-E in parallel, then comparing both results to pick the best one. Use when the user says "cover image", "make a cover", "create a cover image", "cover art", "thumbnail", "hero image", or wants a polished image for a blog post, presentation, video thumbnail, or social media. Takes a text description as the argument.
user-invocable: true
allowed-tools:
  - mcp__claude-in-chrome__tabs_context_mcp
  - mcp__claude-in-chrome__tabs_create_mcp
  - mcp__claude-in-chrome__navigate
  - mcp__claude-in-chrome__read_page
  - mcp__claude-in-chrome__find
  - mcp__claude-in-chrome__form_input
  - mcp__claude-in-chrome__computer
  - mcp__claude-in-chrome__get_page_text
  - mcp__claude-in-chrome__javascript_tool
  - Bash
  - Read
  - Glob
---

# Cover Image — Generate, Compare, Pick the Best

Arguments passed: `$ARGUMENTS`

## Argument Handling

- `$ARGUMENTS` is the concept/topic for the cover image (e.g., "AI agents collaborating", "deep learning neural networks", "startup pitch deck cover")
- If no arguments provided, ASK the user what the cover image should depict

## Workflow

### Step 1: Craft the Prompt

Take the user's concept and craft an image generation prompt that is:

- **Artsy and visually striking** — not generic stock photo vibes
- **NO TEXT in the image** — explicitly instruct both generators to produce zero text, no words, no letters, no typography
- **Abstract or conceptual** — cover images work best when they evoke a feeling rather than depict something literally

Craft TWO prompts — one optimized for each generator's strengths:

- **Gemini prompt**: Lean into photorealistic or cinematic style. Append: "Do not include any text, words, letters, or typography in the image."
- **DALL-E prompt**: Lean into artistic/illustrative style. Append: "No text, no words, no letters, no writing, no typography anywhere in the image."

### Step 2: Generate with Gemini

Invoke the `/gemini` skill with the Gemini-optimized prompt. Wait for the image to be downloaded.

Note the downloaded file path.

### Step 3: Generate with DALL-E

Invoke the `/dall-e` skill with the DALL-E-optimized prompt. Wait for the image to be downloaded.

Note the downloaded file path.

### Step 4: Compare Both Images

Read both downloaded images using the Read tool (it supports image files).

Evaluate each on:
- **Visual impact** — which one grabs attention?
- **Relevance** — which better captures the concept?
- **No-text compliance** — did either accidentally include text?
- **Composition** — which has better balance, color, framing?
- **Cover-worthiness** — which would look better as an actual cover/thumbnail?

### Step 5: Pick the Winner

Present the verdict:

```
Cover Image Comparison
======================

Gemini:  ~/Downloads/<gemini-file>
DALL-E:  ~/Downloads/<dalle-file>

Winner: [Gemini / DALL-E]

Why: [2-3 sentences explaining the choice]

Both images are saved — use the other one if you prefer it.
```

## Rules

- **Always generate NO-TEXT images** — this is non-negotiable. Cover images with AI-generated text look terrible.
- **Both images must be downloaded** before comparison — don't skip one if it's slow.
- **Be honest in the comparison** — don't default to one generator. Judge on merit.
- **If one generator fails**, report the failure and present the successful one as the winner by default.