AgentSkillsCN

header-image-prompt

为博客文章或 GitHub 讨论页头图生成 AI 艺术提示。互动式技能，读取内容、与用户访谈其愿景，并为任何 AI 图像生成工具生成即用型提示。适用于用户说“页头图”、“创建页头”、“为我的帖子生成艺术”等情况。

SKILL.md
--- frontmatter
name: header-image-prompt
description: Generate an AI art prompt for a blog post or GitHub discussion header image. Interactive skill that reads the content, interviews the user about their vision, and produces a ready-to-use prompt for any AI image generation tool. Use when the user says "header image", "create a header", "generate art for my post", or similar.

Header Image Prompt

Generate AI art prompts for blog post and GitHub discussion header images through an interactive interview process.

Overview

This skill reads the content you're creating a header image for, interviews you about your vision, and produces a polished prompt you can paste directly into an AI image generation tool. The output prompt is tool-agnostic -- it works with Nano Banana, Midjourney, DALL-E, or whatever you're using today.

When to Use

Use this skill when:

  • •You're about to publish a blog post and need header art
  • •You're writing a GitHub discussion and want a visual header
  • •You want an illustration-style image that captures the essence of your writing
  • •You have a reference photo you want to incorporate into the concept

Output Specification

All prompts must target these constraints:

  • •Aspect ratio: 5:2 (five units wide, two units tall) -- this is the standard header format
  • •Style default: Illustration (unless the user requests otherwise)
  • •Format: A single prompt block the user can copy-paste into their image generation tool
  • •No tool-specific syntax: Write the prompt as a natural description, not in any tool's proprietary format

Workflow

Step 1: Read the content

Ask the user for the content they want a header image for. This could be:

  • •A file path to a markdown document
  • •A URL to a GitHub discussion or blog post
  • •Pasted text in the conversation

Read the full content. Identify:

  • •The core theme or argument
  • •Key metaphors or imagery already present in the writing
  • •The emotional tone (reflective, energetic, technical, personal)
  • •Any concrete objects, scenes, or concepts that could translate visually

Step 2: Present your initial read

Share a brief summary (2-3 sentences) of what you took away from the content and what visual directions you're already thinking about. This gives the user a chance to correct course early before the interview goes deeper.

Step 3: Interview the user

Ask questions to narrow down the visual direction. Do not ask all questions at once -- have a conversation, adapting based on their answers. Start with the most important questions and follow the thread.

Opening questions (pick 1-2 to start):

  • •What feeling do you want someone to get when they see this image before they start reading?
  • •Is there a specific scene, metaphor, or moment from the writing that you keep picturing?
  • •Are you drawn to something abstract and atmospheric, or something more literal and concrete?

Follow-up questions (use as the conversation develops):

  • •Color palette -- do you have a mood in mind? (warm, cool, muted, vibrant, monochrome)
  • •Should this feel playful, serious, contemplative, energetic?
  • •Any visual elements you definitely want included or excluded?
  • •Have you seen any header images recently that you liked? What about them worked?
  • •Should people appear in the image? If so, how prominent?

Style clarification (if needed):

  • •The default is illustration style. If you want something different (watercolor, line art, flat design, photorealistic, collage, pixel art), let me know.
  • •How detailed vs. minimal? Dense composition or lots of breathing room?

Step 4: Ask about input images

If the user hasn't already mentioned a reference image, ask:

  • •Do you have a photo or image you'd like to use as a starting point or reference? Some ideas:
    • •A photo of yourself to base an illustrated version on
    • •A photo that captures the mood or setting you're going for
    • •A sketch or rough concept you've drawn
    • •A screenshot or diagram from the work itself

If they provide an image, incorporate it into the prompt as a reference description (since the user will need to upload it separately to their image generation tool). Describe how the reference image should influence the output -- as a pose reference, mood reference, composition reference, or style reference.

Step 5: Generate the prompt

Synthesize everything from the interview into a single, well-crafted prompt. Structure it as:

  1. •Subject and scene -- What is depicted
  2. •Style and medium -- Illustration style, technique references
  3. •Composition -- How elements are arranged in the 5:2 frame (wide and short means careful thought about horizontal composition)
  4. •Color and lighting -- Palette and mood
  5. •Atmosphere and detail level -- The overall feeling

Present the prompt in a code block so it's easy to copy.

Important considerations for 5:2 aspect ratio:

  • •This is a very wide, short frame -- avoid tall subjects that would be cropped or tiny
  • •Think panoramic: landscapes, wide scenes, horizontal arrangements
  • •Subjects should be composed to work in this letterbox format
  • •Leave space on the sides -- the image often has text overlaid or sits above a title

Step 6: Refine

After presenting the prompt, ask the user:

  • •Does this capture what you're going for?
  • •Anything you'd add, remove, or change?
  • •Want me to try a completely different direction?

Iterate until they're happy. When refining, present the full updated prompt each time (don't just describe the changes -- give them the complete copy-pasteable prompt).

Tips for Quality Prompts

  • •Be specific about composition in the 5:2 frame -- "a wide panoramic view" or "subject positioned left of center with negative space on the right"
  • •Use concrete visual language, not abstract concepts -- "warm golden hour light filtering through trees" not "a feeling of warmth and nature"
  • •Reference specific illustration styles when helpful -- "in the style of editorial illustration" or "flat vector illustration with subtle gradients"
  • •For images based on reference photos, describe the transformation clearly -- "an illustrated portrait based on the reference photo, rendered in [style], maintaining the pose and expression but stylized with [details]"
  • •Keep prompts between 50-150 words -- long enough to be specific, short enough to not confuse the model

Example Prompts

Start a header image session

code
I need a header image for my latest blog post. Here's the file: Daily Projects/2026-02-13/01 writing with ai guide.md

Quick with context

code
Generate a header image prompt for this discussion I'm about to publish. It's about our team's approach to DDoS mitigation over the past 5 years.

With a reference image

code
I want a header image for my blog post about writing with AI. I have a photo of myself at my desk I'd like to use as a starting point.