Cover Image

Cover Image Generator

Generate blog cover images and Open Graph/social preview images as PNG files. Each cover is a self-contained HTML document with inline CSS, captured as a screenshot via Chrome headless at exactly 1200x630 pixels.

The output is a static PNG suitable for Hugo front matter (image:), HTML ``, and Twitter card metadata. No external images, no JavaScript, no server-side rendering required at capture time.

Overview

Cover images serve as the primary visual identity for a blog post across social platforms, search results, and link previews. Each image is designed from scratch per post, reflecting the topic, tone, and key concepts of the content.

What this skill produces:

• A 1200x630px PNG file rendered from HTML+CSS
• Self-contained HTML with no external image dependencies
• Typography-driven design with CSS decorative elements
• Consistent branding placement

When to use this skill:

• Blog posts need an OG/social preview image
• Hugo front matter needs an image: field
• A series of posts need visually cohesive covers
• Existing covers need updating or redesigning

Detailed Workflow

Step 1: Read the source content

Read the blog post or article to extract the title, key concepts, and overall tone. The cover image must accurately represent the content. Identify:

• The exact title (or a shortened version if it exceeds 40 characters)
• The primary topic category (technical, tutorial, explanatory, announcement)
• Key visual metaphors or concepts from the content
• Whether the post is part of a series (affects series label placement)
• The desired output file path

If the user provides only a title without a file path, ask for the target output location before proceeding.

Step 2: Determine the visual direction

Select a design approach based on the post topic and tone:

• Technical posts (programming, infrastructure, security, blockchain): dark background with bold typography and accent gradients. Conveys depth and technical rigor.
• Tutorial/guide posts (how-to, step-by-step, beginner): light background with clean layout and structured visual hints. Conveys clarity and approachability.
• Conceptual/explanatory posts (architecture, design patterns, theory): gradient background with centered typography. Conveys breadth and sophistication.
• Announcement posts (new feature, series launch, milestone): bold accent colors with prominent typography. Conveys energy and importance.

Consult references/cover-design.md for the full color palette and typography specifications for each direction.

Step 3: Design the HTML+CSS

Create a complete, self-contained HTML file. All styles must be inline within a ` block. No external stylesheets, no tags (except Google Fonts via @import`), no external images, and no JavaScript.

Hard constraints:

• Canvas dimensions: exactly 1200x630px set on body
• overflow: hidden on body to prevent scroll artifacts
• Max 2 fonts: one display font for titles, one sans-serif for metadata
• Google Fonts loaded via @import url(...) inside the `` block
• All decorative elements created with CSS (gradients, shapes, transforms)
• No ` tags, no background-image: url(...)` pointing to external files

See the HTML boilerplate section below for a complete starting template.

Step 4: Write to a temp file

Save the HTML to a temporary location. Use the project's tmp/ directory:

tmp/cover-.html

Derive the slug from the post title (lowercase, hyphens, max 30 characters). Create the tmp/ directory if it does not exist. Writing to tmp/ keeps generated HTML separate from source content and makes cleanup straightforward.

Step 5: Capture the screenshot

Run the screenshot script to render the HTML to PNG:

bash scripts/screenshot.sh tmp/cover-.html /cover.png

The script handles Chrome headless invocation with the correct flags (--headless=new, --disable-gpu, --no-sandbox, --hide-scrollbars). It outputs the PNG at the specified path.

Important: Run the script from the skill directory so that scripts/screenshot.sh resolves correctly, or use an absolute path to the script.

Step 6: Verify the output

Read the generated PNG file to visually verify it rendered correctly. Check for:

• Title text is fully visible and not clipped
• Text is readable at half-size (simulate by checking overall layout density)
• Decorative elements are positioned correctly (not overlapping critical text)
• Colors render as specified (Chrome may handle some CSS differently)
• No white/blank areas where content should be
• Safe margins are respected (no text too close to edges)

If the image does not look correct, adjust the HTML and re-run the screenshot. Common fixes are in the troubleshooting section below.

Step 7: Clean up

After confirming the PNG renders correctly, delete the temp HTML file:

rm tmp/cover-.html

Do not leave temp HTML files in the repository. If the tmp/ directory is now empty, remove it as well.

Step 8: Integrate with the blog post

Add the cover image to the blog post front matter:

---
title: "Post Title"
image: /images/category/cover.png
---

Copy or move the PNG to the appropriate static images directory for the blog. The exact path depends on the blog's static file structure.

HTML Boilerplate

Use this template as the starting point for every cover image. It includes the correct dimensions, font loading, layout zones, and branding placement.

Dark Variant Template

  @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;600;700;800&family=JetBrains+Mono:wght@400;700&display=swap');

  * { margin: 0; padding: 0; box-sizing: border-box; }

  body {
    width: 1200px;
    height: 630px;
    overflow: hidden;
    font-family: 'Inter', 'Helvetica Neue', Arial, sans-serif;
    background: #0f172a;
    color: #f8fafc;
    position: relative;
  }

  /* Background decoration */
  .bg-circle {
    position: absolute;
    border-radius: 50%;
    opacity: 0.06;
  }
  .bg-circle-1 {
    width: 600px;
    height: 600px;
    background: linear-gradient(135deg, #6366f1, #8b5cf6);
    top: -200px;
    right: -100px;
  }
  .bg-circle-2 {
    width: 400px;
    height: 400px;
    background: linear-gradient(135deg, #06b6d4, #3b82f6);
    bottom: -150px;
    left: -50px;
  }

  /* Layout */
  .cover {
    position: relative;
    z-index: 1;
    padding: 60px 80px;
    height: 100%;
    display: flex;
    flex-direction: column;
    justify-content: center;
  }

  /* Series label */
  .series {
    font-size: 14px;
    font-weight: 600;
    text-transform: uppercase;
    letter-spacing: 1.5px;
    color: #818cf8;
    margin-bottom: 20px;
  }

  /* Title */
  .title {
    font-size: 56px;
    font-weight: 800;
    line-height: 1.15;
    max-width: 900px;
    margin-bottom: 16px;
  }

  /* Subtitle */
  .subtitle {
    font-size: 22px;
    font-weight: 400;
    color: #94a3b8;
    max-width: 700px;
    line-height: 1.4;
  }

  /* Branding */
  .branding {
    position: absolute;
    bottom: 24px;
    right: 80px;
    font-size: 14px;
    color: #94a3b8;
    font-weight: 500;
  }

  
  

  
    Series Label
    Post Title Goes Here
    Optional subtitle or tagline
  

  widnyana.web.id

Light Variant Template

Replace the dark-specific styles:

body {
  background: #ffffff;
  color: #0f172a;
}

.bg-circle {
  opacity: 0.04;
}

.series {
  color: #3b82f6;
}

.subtitle {
  color: #64748b;
}

.branding {
  color: #64748b;
}

The structure remains identical. Only background, text, and accent colors change.

Design Rules

When to use dark vs light

| Topic Type | Background | Rationale | |------------|-----------|-----------| | Programming, Rust, Solidity | Dark (#0f172a) | Technical depth | | Infrastructure, DevOps, CI/CD | Dark (#0f172a) | Systems-level content | | Security, auditing | Dark (#0f172a) | Seriousness, gravity | | Blockchain, crypto, DeFi | Dark (#0f172a) | Industry convention | | Tutorials, guides, how-to | Light (#ffffff) | Approachable, clear | | Architecture, design patterns | Gradient | Breadth, sophistication | | Career, soft skills | Light (#ffffff) | Friendly, readable | | Announcements, launches | Bold gradient | Energy, prominence |

Font loading

Always load fonts via @import inside the ` block, not via in the head. This keeps the HTML truly self-contained and avoids rendering issues where Chrome headless may not wait for external ` resources.

Include only the weights actually used. Loading unused weights slows Chrome headless rendering:

/* Good: only needed weights */
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;700;800&display=swap');

/* Bad: loading all weights */
@import url('https://fonts.googleapis.com/css2?family=Inter:wght@100..900&display=swap');

Dimensions and safe area

The full canvas is 1200x630px. Social platforms (Twitter, LinkedIn, Slack, Discord) may crop edges. Critical content (title text, series labels) must stay within the safe area:

• Safe area: 1040 x 510px centered within the canvas
• Left/right margin: minimum 80px
• Top/bottom margin: minimum 60px
• Branding: stays within safe area (bottom: 20-50px), uses solid muted colors with >= 5.5:1 contrast. Omit entirely when no branding is provided.

Visual element guidelines

Each cover must include at least one visual element beyond pure typography. Options:

1. Gradient orbs: Large translucent circles positioned behind the text area. Use 2-3 orbs with low opacity (0.04-0.08) for depth.
2. Accent bars: Thin horizontal or diagonal lines with gradient colors to add structure.
3. Geometric shapes: Triangles, rotated squares, or hexagons as subtle background patterns.
4. Code snippet visuals: For programming posts, a styled code block with syntax-colored text adds context.
5. Grid/dot patterns: Repeating dot or line patterns using CSS background-image with linear-gradient for technical posts.

All decorative elements must use position: absolute and sit behind text content via z-index. Never let decorations interfere with text legibility.

Common Issues and Troubleshooting

Chrome not found

Symptom: scripts/screenshot.sh exits with error: no Chrome/Chromium binary found in PATH.

Fix: Install Chrome or Chromium. The script searches for google-chrome-stable, google-chrome, chromium-browser, or chromium in that order. On Fedora: sudo dnf install chromium. On Ubuntu: sudo apt install chromium-browser. Verify with which google-chrome-stable.

Font not loading or rendering as fallback

Symptom: Title renders in Arial or Times New Roman instead of the specified Google Font.

Fixes:

• Verify the @import URL is correct and the font name matches exactly (case-sensitive).
• Ensure the @import is the first rule inside `. CSS @import` must precede all other rules.
• Add a small font-display: swap parameter: &display=swap in the Google Fonts URL.
• If the font still fails, fall back to a web-safe stack: 'Inter', 'Helvetica Neue', Arial, sans-serif.

Text clipped or overflowing

Symptom: Title or subtitle text is cut off at the bottom or right edge.

Fixes:

• Check that overflow: hidden is set on body (this is intentional, but confirms clipping is happening).
• Reduce font size if the title exceeds 2 lines at the current size.
• Increase max-width on the title element, but never exceed 1000px to preserve safe margins.
• Shorten the title text for the cover. The cover title does not need to match the full post title exactly.

Image appears blank or white

Symptom: The output PNG is entirely white or appears empty.

Fixes:

• Verify the HTML file exists and is not empty before running the screenshot script.
• Check that body has an explicit background color set. Chrome headless defaults to white.
• Ensure all content has z-index higher than 0 if background elements are present.
• Open the HTML file in a regular browser to confirm it renders correctly before blaming headless Chrome.

Colors look different from CSS values

Symptom: Gradient or background colors appear washed out or shifted compared to the CSS specification.

Fixes:

• Chrome headless renders in sRGB color space. If the display preview uses a wider gamut, colors may appear different.
• Avoid color() or lab() CSS color functions. Use hex (#0f172a) or rgba() values only.
• Verify the PNG is not being displayed at a different size than 1200x630 (scaling changes color perception).

Screenshot shows scrollbars or extra whitespace

Symptom: The PNG has a scrollbar or extra whitespace on the right or bottom.

Fixes:

• Confirm overflow: hidden is set on body.
• Add margin: 0; padding: 0 to both html and body via the universal selector * { margin: 0; padding: 0; }.
• The screenshot script passes --hide-scrollbars but this may not suppress all scrollbar artifacts in all Chrome versions. The CSS overflow: hidden is the reliable fix.

Quick Reference

Cover Image Specifications

| Property | Value | |----------|-------| | Canvas size | 1200 x 630 px | | Safe area | 1040 x 510 px (centered) | | Format | PNG | | Min margins | 80px horizontal, 60px vertical | | Max fonts | 2 (display + body) | | External images | None (CSS only) | | JavaScript | None (static HTML+CSS) |

Output Locations

| Context | Typical Path | |---------|-------------| | Hugo blog | static/images//cover.png | | Temp HTML | tmp/cover-.html |

Color Palettes (Quick)

| Variant | Background | Primary Text | Secondary Text | Accent | |---------|-----------|-------------|---------------|--------| | Dark | #0f172a | #f8fafc | #94a3b8 | #6366f1 to #8b5cf6 | | Light | #ffffff | #0f172a | #64748b | #3b82f6 | | Gradient | #1e1b4b to #312e81 | #f8fafc | #a5b4fc | #818cf8 |

Workflow Checklist

1. Read the post content
2. Select dark/light/gradient direction
3. Write self-contained HTML with inline CSS
4. Save to tmp/cover-.html
5. Run bash scripts/screenshot.sh tmp/cover-.html .png
6. Read the PNG to verify rendering
7. Delete temp HTML
8. Add to blog front matter

Additional Resources

Local References

• references/cover-design.md - Complete color palettes, typography scale, layout zones, and visual element specifications for cover images
• scripts/screenshot.sh - Chrome headless capture script that handles binary detection, viewport sizing, and output file management

Related Design Guides

For diagram and infographic styles used in other visual-gen skills, consult the architecture-diagram and process-infographic skill references.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: widnyana
• Source: widnyana/eyay-toolkits
• License: MIT

Install and usage instructions live in the source repository linked above.