Agent infographic skill
Create original, evidence-led editorial infographics and social graphics that avoid generic AI aesthetics. Use for infographics, news cards, data stories, Pinterest pins, social carousels, visual explainers, diagrams, or when a graphic must explain, prove, compare, map, or humanize an idea.From its SKILL.md
npx -y skills add aosoficial/agent-infographic-skillAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
4.7 KB, 949 tokens by cl100k_base, as published. Nobody here has run it
Editorial Infographics
Create visuals that communicate before they decorate. The output should feel art-directed, not generated from a generic social template.
Non-negotiable outcome
Every visual must perform one primary job:
- Explain a process or mechanism.
- Prove a claim with a real number, quote, screenshot, or source.
- Compare two states with an honest common scale.
- Map relationships, hierarchy, overlap, or flow.
- Humanize an abstract idea with a relevant person, object, or situation.
If the visual only restates the post title, stop and redesign it.
Workflow
1. Build an evidence pack
Collect the source text and any real data, screenshots, quotes, product UI, photos, dates, or documents that support it. Mark each claim as verified, inference, or unknown.
Never invent statistics, interfaces, quotes, sources, or product behavior. Capture real UI instead of generating fake UI.
2. Write the one-sentence takeaway
Complete this sentence:
After seeing this image for five seconds, the viewer should understand that …
One image gets one takeaway. Split a dense idea into a carousel or series.
3. Classify the information topology
Choose structure before style:
- progression or convergence →
convergence - why / how / what or nested scope →
concentric - levels, maturity, risk, or containment →
layers - shared causes, systems, or intersections →
overlap - before/after →
binary-comparison - a measured trend →
annotated-chart - product proof →
real-screenshot
Read layout families before selecting a composition.
4. Art-direct the visual
Use the design language. Start from a quiet editorial canvas, one dominant diagram, one accent color, and a deliberate asymmetry. Typography and geometry should carry the design.
Do not imitate a reference literally. Extract its hierarchy, rhythm, spatial metaphor, density, and contrast, then create a new composition.
5. Choose a production method
Prefer the method with the highest truth and control:
- Real screenshot or photo with restrained annotation.
- Deterministic vector, HTML/CSS, chart, or diagram for exact data and text.
- AI-generated illustration only when a metaphor genuinely adds meaning.
- Hybrid: generated background or object plus deterministic type and labels.
Do not ask an image model to invent dense body copy or precise charts. Render text and data deterministically.
6. Adapt to the platform
Create a native composition for each aspect ratio. Never stretch one master file. Read platform formats.
If the asset ships with social copy, run the publishing transport gate. Paragraph spacing must survive the scheduler: visible \\n, raw HTML, collapsed paragraphs, or duplicated queue items are release blockers.
7. Run the anti-slop gate
Score the piece with quality gate. A visual cannot ship if:
- the visual job is unclear;
- its main claim lacks evidence or is framed as fact without support;
- it repeats the same layout as another item in the batch;
- the title is the only useful information;
- it contains fake UI, fake data, illegible text, or decorative complexity;
- it resembles a generic AI template more than a specific editorial decision.
- the scheduled copy contains literal escape sequences, unexpected HTML, or broken paragraph spacing.
8. Preserve a receipt
Save the brief, sources, final copy, dimensions, production method, file hash, and QA score. When publishing a batch, record which layout family each asset used.
Batch diversity rule
For any batch of four or more visuals:
- use at least three layout families;
- no more than two assets may share the same dominant composition;
- alternate density and scale;
- include at least one evidence-led visual;
- include at least one visual with no card grid;
- inspect the batch as thumbnails before publishing.
Required output
Deliver:
- Evidence pack and source status.
- One-sentence takeaway.
- Chosen visual job and layout family.
- Platform dimensions.
- Final overlay copy.
- Production method and reproducible prompt or spec.
- Final asset path.
- QA score and any limitations.
Use the visual brief template as the working record.
What ships with it: 10 files
18.2 KB alongside SKILL.md
examples/
- cfo-capacity/brief.md1.5 KB
references/
- design-language.md2.3 KB
- layout-families.md2.5 KB
- platforms.md1.3 KB
- publishing-gate.md1.4 KB
- quality-gate.md1.9 KB
- research.md2.1 KB
templates/
- visual-brief.md903 B