blog-cover-image
Skill2 starsChanged 3 months agoApache-2.0View on GitHub
Generate and revise raster cover images for blog posts and articles from a title, brief, draft, local file, URL content supplied by the user, or reference images. Use when Codex needs to create a blog cover, article hero/header image, editorial thumbnail, Open Graph/social preview visual, or a channel-specific cover variant, including selecting composition, palette, rendering, text density, mood, font style, and aspect ratio.
3 other projects use a version of this.
Adopted by
Most recently taken up by
- alfredang/ai-cmstheir own version, first seen Nov 7, 2026
- alfredang/ai4kidstheir own version, first seen Nov 2, 2026
- alfredang/ai-peitheir own version, first seen Jun 12, 2026
Other versions3 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 3
- alfredang/ai-cms★ 4same name
- alfredang/ai-pei★ 1same name
- alfredang/ai4kids★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository4 other files it gives its agents.
- skills/humanizer/AGENTS.mdAGENTS.md
- content-strategySkill
- groniz-cliSkill
- humanizerSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
31 rules. 0 are common, 29 are distinctive, and 2 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
29
Distinctive
This author's own choices. Worth a look.
2
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
Create a clear, content-led raster cover that works at its intended display size.
- Distinctive: line 12
Read the supplied content before designing.
- Distinctive: line 15
Use an available raster image-generation or image-editing capability.
- Distinctive: line 16
Do not substitute SVG, HTML, CSS, canvas, or another code-rendered visual unless the user explicitly requests that output.
- Distinctive: line 17
If no raster backend is available, report the limitation and ask the user how to proceed.
- Distinctive: line 18
Do not alter the source article while creating its cover.
- Distinctive: line 19
Do not patch misspelled or garbled generated text by painting over the bitmap.
- Distinctive: line 32
Prefer explicit user choices.
- Distinctive: line 34
Use these defaults when no better signal exists:
- Distinctive: line 48
Use requested pixel dimensions when provided.
- Distinctive: line 52
Use six independent dimensions.
- Distinctive: line 63
Read auto-selection.md when one or more dimensions are unspecified.
- Distinctive: line 72
Treat compatibility tables as recommendations, not prohibitions.
- Distinctive: line 88
Prefer a meaningful focal idea over a collage of literal keywords.
- Very specific: line 92
If reference images are supplied, inspect them before composing the prompt.
- Distinctive: line 102
Use the exact supplied title.
- Very specific: line 103
Prefer none when the title is long, the destination already renders the headline, or the backend is unreliable with the required script.
- Distinctive: line 104
Keep a title short enough to remain legible at thumbnail size.
- Distinctive: line 105
Do not invent a subtitle merely to fill space.
- Distinctive: line 109
Read base-prompt.md and prompt-template.md.
- Distinctive: line 120
When working in a filesystem and an output directory is in scope, save the final prompt before generation:
- Distinctive: line 130
Use a two-to-four-word lowercase kebab-case slug.
- Distinctive: line 136
Use the backend explicitly requested by the user.
- Distinctive: line 137
Prefer the current runtime's native raster image capability.
- Distinctive: line 138
Use the only available compatible raster backend.
- Distinctive: line 139
If multiple materially different backends remain, ask the user to choose.
- Distinctive: line 151
No unintended words, watermarks, logos, UI fragments, extra limbs, or irrelevant objects
- Distinctive: line 155
If a required check fails, preserve the flawed candidate, strengthen the prompt, and regenerate once.
- Distinctive: line 170
Create a new prompt record before each regeneration.
- Distinctive: line 171
Keep the core visual identity stable across channel variants.
- Distinctive: line 172
Use image editing for localized visual changes when the active backend supports it.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Blog Cover Image
Create a clear, content-led raster cover that works at its intended display size.
Guardrails
- Read the supplied content before designing. Treat its title, subject, tone, and claims as authoritative.
- Preserve the exact user-provided or source title. Never invent a title, subtitle, logo, endorsement, product UI, statistic, or brand element.
- Honor explicit brand guidelines, reference images, dimensions, file format, safe areas, and output paths before applying this skill's defaults.
- Use an available raster image-generation or image-editing capability. Read and follow that capability's own instructions before invoking it.
- Do not substitute SVG, HTML, CSS, canvas, or another code-rendered visual unless the user explicitly requests that output.
- If no raster backend is available, report the limitation and ask the user how to proceed.
- Do not alter the source article while creating its cover.
- Do not patch misspelled or garbled generated text by painting over the bitmap. Regenerate with a stronger prompt or remove in-image text.
Resolve Inputs
Collect or infer:
- Content source: title, topic, brief, draft, or article path
- Intended placement or channel
- Output location and format
- Aspect ratio or pixel dimensions
- Text to render, if any
- Brand constraints and reference images
Prefer explicit user choices. Infer omitted choices when the content and destination make the answer clear; ask only when an unresolved choice would materially change the result.
Use these defaults when no better signal exists:
| Setting | Default |
| Aspect ratio | 16:9 |
| Type | Auto-select from content |
| Palette | Auto-select from content |
| Rendering | Auto-select from content |
| Text | title-only for a short supplied title; otherwise none |
| Mood | balanced |
| Font | clean |
| Language | Match the source or user request |
| Format | PNG |
Use requested pixel dimensions when provided. Otherwise treat aspect ratio as the contract and let the active raster backend choose supported dimensions.
Design Dimensions
Use six independent dimensions. A style preset is only shorthand for palette and rendering; explicit settings override it.
| Dimension | Values |
| Type | hero, conceptual, typography, metaphor, scene, minimal |
| Palette | warm, elegant, cool, dark, earth, vivid, pastel, mono, retro, duotone, macaron |
| Rendering | flat-vector, hand-drawn, painterly, digital, pixel, chalk, screen-print |
| Text | none, title-only, title-subtitle, text-rich |
| Mood | subtle, balanced, bold |
| Font | clean, handwritten, serif, display |
Read auto-selection.md when one or more dimensions are unspecified. Read only the selected dimension references needed to build the final prompt:
- types.md for composition type
- palettes/ for the selected palette
- renderings/ for the selected rendering
- text.md, mood.md, and font.md for the remaining dimensions
- compatibility.md when combining unusual choices
- style-presets.md when the user names a preset
Treat compatibility tables as recommendations, not prohibitions.
Workflow
1. Analyze the Content
Identify:
- Exact title and optional subtitle
- Central idea in one sentence
- Audience and intended outcome
- Tone and emotional intensity
- Three to eight concrete keywords
- One strong visual metaphor or focal subject
- Visuals to avoid because they would be misleading, generic, or off-brand
Prefer a meaningful focal idea over a collage of literal keywords.
2. Inspect References
If reference images are supplied, inspect them before composing the prompt. Read reference-images.md and classify each reference as direct, style, palette, or composition guidance.
Describe observable traits such as layout, colors, line quality, texture, lighting, subject placement, and typography. Do not rely on vague instructions such as "make it similar."
3. Select a Cover Specification
Choose the six dimensions, aspect ratio, language, text content, and output format. Use option-selection.md when choices are missing or conflicting.
For in-image text:
- Use the exact supplied title.
- Prefer
nonewhen the title is long, the destination already renders the headline, or the backend is unreliable with the required script. - Keep a title short enough to remain legible at thumbnail size.
- Do not invent a subtitle merely to fill space.
4. Build a Reproducible Prompt
Read base-prompt.md and prompt-template.md. Write one self-contained prompt containing:
- Content context
- Exact text, or an explicit instruction to render no text
- Aspect ratio and intended placement
- Type and spatial composition
- Palette with concrete colors
- Rendering, mood, and typography treatment
- Required reference-image traits
- Negative constraints
When working in a filesystem and an output directory is in scope, save the final prompt before generation:
<output-dir>/
├── prompts/
│ └── 01-cover-<slug>.md
├── refs/ # only when references are persisted
└── cover.<ext>Use a two-to-four-word lowercase kebab-case slug. Preserve existing candidates by incrementing the prompt and output filenames for revisions.
5. Generate the Raster Image
Select the backend in this order:
- Use the backend explicitly requested by the user.
- Prefer the current runtime's native raster image capability.
- Use the only available compatible raster backend.
- If multiple materially different backends remain, ask the user to choose.
Pass the final prompt, requested aspect ratio or dimensions, and every direct reference through the backend's supported interface. If the backend manages output paths, report that path; if it supports a target path, write to the agreed location.
6. Inspect and Validate
Inspect the generated image whenever the runtime permits it. Verify:
- Correct aspect ratio and usable crop
- One clear focal point and readable hierarchy
- Accurate, legible, exact text
- Adequate contrast and safe margins
- No unintended words, watermarks, logos, UI fragments, extra limbs, or irrelevant objects
- Required reference and brand traits are visible
- The image still communicates at thumbnail size
If a required check fails, preserve the flawed candidate, strengthen the prompt, and regenerate once. Ask for direction after repeated subjective failures; continue iterating automatically on objective failures such as wrong text or dimensions when the backend permits.
7. Report Completion
Return:
- Final image or its path
- Type, palette, rendering, text, mood, font, and aspect ratio
- Prompt path when saved
- Reference count
- Any limitation that remains, especially imperfect generated text
Revisions and Variants
- Preserve the original cover unless the user explicitly asks to overwrite it.
- Create a new prompt record before each regeneration.
- Keep the core visual identity stable across channel variants; adjust crop, safe area, text density, and composition for each aspect ratio.
- Use image editing for localized visual changes when the active backend supports it; regenerate when the requested change affects composition or text.
- Limit post-processing to crop, resize, compression, and format conversion unless the user explicitly requests a composite workflow.
Reference Index
- Selection: auto-selection.md, compatibility.md, style-presets.md
- Composition: types.md, visual-elements.md
- Prompting: base-prompt.md, prompt-template.md
- References: reference-images.md
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

