Baoyu Infographic Skill for Hermes Agent: Layouts, Styles, Workflow
Infographics: 21 layouts x 21 styles (信息图, 可视化).
Written by Neura Market from the official Hermes Agent documentation for Baoyu Infographic. Commands, paths, and version numbers are reproduced from the source unchanged.
Read the official documentationThe Baoyu Infographic skill turns a block of text, a file, a URL, or a topic into a single generated infographic image. It is bundled with Hermes Agent, so it is already available in a default install. Reach for it when you need a visual summary that keeps the source numbers and wording intact, and you want control over how the information is arranged and how it looks.
What it does
The skill works along two independent axes. The first is layout, which decides how information is structured on the page: a timeline, a comparison, a funnel, a dashboard, and so on. The second is style, which decides the visual treatment: hand-drawn paper craft, neon cyberpunk, blueprint schematic, retro pop, and so on. Any layout can be paired with any style, so the combination space is large.
In practice the skill runs a short pipeline. It saves your source material, analyses it, restructures it into a markdown outline with learning objectives and verbatim data points, recommends a few layout and style pairings, confirms the choices with you, assembles a detailed image prompt, and then calls the image generation tool. The output is a folder containing the source, the analysis, the structured content, the prompt, and the final image.
The author is 宝玉 (JimLiu), the licence is MIT, and the skill is adapted from the upstream baoyu-skills project for Hermes Agent's tool ecosystem. It runs on linux, macos, and windows.
Before you start
The skill is bundled, so there is nothing to install. Its path inside the runtime is skills/creative/baoyu-infographic, and the version is 1.56.1.
Two practical prerequisites follow from how the workflow is built. First, the final step calls the image_generate tool, so that tool needs to be available in your Hermes Agent setup for an image to be produced. Second, the skill reads several reference files from its own directory as it runs: references/analysis-framework.md, references/structured-content-template.md, references/base-prompt.md, references/layouts/.md, and references/styles/.md. These ship with the skill.
You supply the content. It can be text you paste, a file path, a URL, or just a topic. You can also state a layout, a style, an aspect ratio, or a language up front, but none of these are required.
Options
Four settings shape the result. Layout and style each have 21 choices. Aspect can be a named preset or a custom ratio. Language is open, with English, Chinese, Japanese, and others as examples.
| Option | Values |
|---|---|
| Layout | 21 options (see Layout Gallery), default: bento-grid |
| Style | 21 options (see Style Gallery), default: craft-handmade |
| Aspect | Named: landscape (16:9), portrait (9:16), square (1:1). Custom: any W:H ratio (e.g., 3:4, 4:3, 2.35:1) |
| Language | en, zh, ja, etc. |
If you say nothing, you get bento-grid with craft-handmade.
Layout Gallery
Layout is the structural decision, and it matters more than style for whether the infographic communicates. Pick the layout that matches the shape of your content, not the one that looks most interesting.
| Layout | Best For |
|---|---|
linear-progression | Timelines, processes, tutorials |
binary-comparison | A vs B, before-after, pros-cons |
comparison-matrix | Multi-factor comparisons |
hierarchical-layers | Pyramids, priority levels |
tree-branching | Categories, taxonomies |
hub-spoke | Central concept with related items |
structural-breakdown | Exploded views, cross-sections |
bento-grid | Multiple topics, overview (default) |
iceberg | Surface vs hidden aspects |
bridge | Problem-solution |
funnel | Conversion, filtering |
isometric-map | Spatial relationships |
dashboard | Metrics, KPIs |
periodic-table | Categorized collections |
comic-strip | Narratives, sequences |
story-mountain | Plot structure, tension arcs |
jigsaw | Interconnected parts |
venn-diagram | Overlapping concepts |
winding-roadmap | Journey, milestones |
circular-flow | Cycles, recurring processes |
dense-modules | High-density modules, data-rich guides |
Full definitions live in references/layouts/.md.
Style Gallery
Style is the aesthetic layer. It changes how the same structure reads: playful, serious, technical, nostalgic. The default is craft-handmade.
| Style | Description |
|---|---|
craft-handmade | Hand-drawn, paper craft (default) |
claymation | 3D clay figures, stop-motion |
kawaii | Japanese cute, pastels |
storybook-watercolor | Soft painted, whimsical |
chalkboard | Chalk on black board |
cyberpunk-neon | Neon glow, futuristic |
bold-graphic | Comic style, halftone |
aged-academia | Vintage science, sepia |
corporate-memphis | Flat vector, vibrant |
technical-schematic | Blueprint, engineering |
origami | Folded paper, geometric |
pixel-art | Retro 8-bit |
ui-wireframe | Grayscale interface mockup |
subway-map | Transit diagram |
ikea-manual | Minimal line art |
knolling | Organized flat-lay |
lego-brick | Toy brick construction |
pop-laboratory | Blueprint grid, coordinate markers, lab precision |
morandi-journal | Hand-drawn doodle, warm Morandi tones |
retro-pop-grid | 1970s retro pop art, Swiss grid, thick outlines |
hand-drawn-edu | Macaron pastels, hand-drawn wobble, stick figures |
Full definitions live in references/styles/.md.
Recommended combinations
The skill ships a table of pairings that are known to work well together. Treat it as a starting point, not a rule. If your content is a timeline, linear-progression with craft-handmade is the suggested pairing. If it is a step-by-step tutorial, linear-progression with ikea-manual reads more clearly.
| Content Type | Layout + Style |
|---|---|
| Timeline/History | linear-progression + craft-handmade |
| Step-by-step | linear-progression + ikea-manual |
| A vs B | binary-comparison + corporate-memphis |
| Hierarchy | hierarchical-layers + craft-handmade |
| Overlap | venn-diagram + craft-handmade |
| Conversion | funnel + corporate-memphis |
| Cycles | circular-flow + craft-handmade |
| Technical | structural-breakdown + technical-schematic |
| Metrics | dashboard + corporate-memphis |
| Educational | bento-grid + chalkboard |
| Journey | winding-roadmap + storybook-watercolor |
| Categories | periodic-table + bold-graphic |
| Product Guide | dense-modules + morandi-journal |
| Technical Guide | dense-modules + pop-laboratory |
| Trendy Guide | dense-modules + retro-pop-grid |
| Educational Diagram | hub-spoke + hand-drawn-edu |
| Process Tutorial | linear-progression + hand-drawn-edu |
Default: bento-grid + craft-handmade
Keyword shortcuts
Some phrases in your request trigger an automatic layout choice. When the input contains one of these keywords, the skill skips content-based layout inference and selects the associated layout. It then offers the associated styles as the top recommendations during the confirmation step. If a shortcut has prompt notes, those notes are appended to the generated prompt as extra style instructions.
| User Keyword | Layout | Recommended Styles | Default Aspect | Prompt Notes |
|---|---|---|---|---|
| 高密度信息大图 / high-density-info | dense-modules | morandi-journal, pop-laboratory, retro-pop-grid | portrait | , |
| 信息图 / infographic | bento-grid | craft-handmade | landscape | Minimalist: clean canvas, ample whitespace, no complex background textures. Simple cartoon elements and icons only. |
This is useful when you already know what you want. Saying "high-density info" commits you to dense-modules and a portrait orientation, which is the right call for a data-rich guide but wrong for a single concept.
Output structure
Every run produces a folder named after the topic. The slug is two to four kebab-case words derived from the topic. If a folder with that name already exists, the skill appends a timestamp in the form -YYYYMMDD-HHMMSS.
infographic/{topic-slug}/
├── source-{slug}.{ext}
├── analysis.md
├── structured-content.md
├── prompts/infographic.md
└── infographic.png
Keeping the intermediate files matters. If the image comes out wrong, you can inspect structured-content.md to see whether the content was restructured badly, or prompts/infographic.md to see whether the layout and style definitions were assembled correctly, and rerun from there.
Core principles
The skill enforces three rules throughout. Source data is preserved faithfully, with no summarisation or rephrasing, though any credentials, API keys, tokens, or secrets are stripped before they reach an output file. Learning objectives are defined before the content is structured. And the structure is built for visual communication, using headlines, labels, and visual elements rather than paragraphs.
Workflow
Step 1: Analyse content
The skill first reads references/analysis-framework.md from its own directory. It then saves your source content to source.md using write_file. If source.md already exists, it is renamed to source-backup-YYYYMMDD-HHMMSS.md first. The analysis covers topic, data type, complexity, tone, and audience, and it detects both the source language and your language. Any design instructions in your input are extracted. The result is written to analysis.md, with the same backup rule applied to any existing file.
Step 2: Generate structured content
The content is transformed into an infographic structure in structured-content.md: a title and learning objectives, then sections each carrying a key concept, the verbatim content, a visual element, and text labels. All statistics and quotes are copied exactly. Your design instructions are carried through. The file is markdown only, adds no new information, and has secrets stripped.
The detailed format is described in references/structured-content-template.md.
Step 3: Recommend combinations
The skill checks the keyword shortcuts first. If your input matches one, the associated layout is selected automatically and the associated styles become the top recommendations, with content-based layout inference skipped. Otherwise it recommends three to five layout and style combinations based on data structure, content tone, audience expectations, and your design instructions.
Step 4: Confirm options
The clarify tool is used to confirm the choices with you. Independent questions go into a single questions array, up to five. Any question whose options depend on an earlier answer is asked separately.
Q1 asks you to pick from three or more layout and style combinations, each with a rationale. Q2 asks for an aspect ratio: landscape, portrait, square, or a custom W:H. Q3 asks which language the text content should use, and only appears when the source language differs from your language.
Step 5: Generate the prompt
The selected layout definition is read from references/layouts/.md and the selected style from references/styles/.md. These are combined with the base template from references/base-prompt.md, the structured content from Step 2, and all text in the confirmed language. The assembled prompt is saved to prompts/infographic.md using write_file, with the same backup rule applied to any existing file.
Aspect ratio resolution works like this. Named presets become ratio strings: landscape becomes 16:9, portrait becomes 9:16, square becomes 1:1. Custom W:H ratios are used as-is, for example 3:4, 4:3, or 2.35:1.
Step 6: Generate the image
The image_generate tool is called with the assembled prompt. The aspect ratio is mapped to the tool's format: 16:9 becomes landscape, 9:16 becomes portrait, and 1:1 becomes square. For custom ratios, the closest named aspect is chosen. If generation fails, it retries once automatically. The resulting image URL or path is saved to the output directory.
Step 7: Output summary
The skill reports the topic, layout, style, aspect, language, output path, and the files created.
When not to use it
The source does not describe alternatives or exclusions. The trigger conditions are specific: an infographic, visual summary, information graphic, or one of the Chinese terms 信息图, 可视化, or 高密度信息大图. If you need a plain text summary or a diagram that must be edited later, this skill produces a single rendered image, so it is not the right tool for those jobs.
Limits and gotchas
The source lists five pitfalls, and they are worth reading before your first run.
- Data integrity is paramount. Never summarise, paraphrase, or alter source statistics. "73% increase" must stay "73% increase", not "significant increase".
- Strip secrets. Always scan source content for API keys, tokens, or credentials before including it in any output file.
- One message per section. Each infographic section should convey one clear concept. Overloading sections reduces readability.
- Style consistency. The style definition from the references file must be applied consistently across the entire infographic. Do not mix styles.
- image_generate aspect ratios. The tool only supports
landscape,portrait, andsquare. Custom ratios like3:4should map to the nearest option, which is portrait in that case.
There is also a practical consequence of the backup rule. Repeated runs on the same topic accumulate timestamped backup files rather than overwriting, so a long session leaves a trail of source-backup-*, analysis-backup-*, and prompts/infographic-backup-* files in the output folder.
What pairs with this
The skill depends on the image_generate tool for its final step and on the clarify tool for option confirmation. Its reference files are internal to the skill directory. The source page does not list other skills that pair with it.
References
references/analysis-framework.md, Analysis methodologyreferences/structured-content-template.md, Content formatreferences/base-prompt.md, Prompt templatereferences/layouts/.md, 21 layout definitionsreferences/styles/.md, 21 style definitions