Baoyu Comic Skill for Hermes Agent: Create Knowledge Comics from Any Content
Knowledge comics (知识漫画): educational, biography, tutorial.
Written by Neura Market from the official Hermes Agent documentation for Baoyu Comic. Commands, paths, and version numbers are reproduced from the source unchanged.
Read the official documentationNeura Market Comic Creation Tool Reference
Overview
This tool creates original knowledge and educational comics from user-provided content. It supports flexible combinations of art styles and tones, and is adapted from the baoyu-comic project for the Hermes Agent tool ecosystem. The tool produces a structured output directory containing analysis, storyboard, character definitions, prompts, and final page images.
When to Use
Use this tool when a user asks to create a knowledge or educational comic, a biography comic, a tutorial comic, or uses terms like "知识漫画", "教育漫画", or "Logicomix-style". The user provides content (text, file path, URL, or topic) and may optionally specify art style, tone, layout, aspect ratio, or language.
Prerequisites
- Hermes Agent environment with
image_generatetool available. - Access to
curlor equivalent for downloading images from URLs. - Ability to write files to
comic/{topic-slug}/directory structure. - Reference files (
analysis-framework.md,character-template.md,storyboard-template.md,art-styles/,tones/,presets/,layouts/,workflow.md,auto-selection.md,partial-workflows.md,ohmsha-guide.md) available at specified paths. - User provides content (text, file path, URL, or topic) and optionally style/tone/layout/aspect/language preferences.
Capabilities
- Analyze source content and save to
source-{slug}.md. - Extract style, palette, or scene traits from user-supplied reference images (text only, not passed to image model).
- Support 6 art styles:
ligne-claire,manga,realistic,ink-brush,chalk,minimalist. - Support 7 tones:
neutral,warm,dramatic,romantic,energetic,vintage,action. - Support 5 presets with special rules:
ohmsha,wuxia,shoujo,concept-story,four-panel. - Support 7 layouts: standard, cinematic, dense, splash, mixed, webtoon, four-panel.
- Support 3 aspect ratios: 3:4 (portrait), 4:3 (landscape), 16:9 (widescreen).
- Support language auto-detection or explicit setting (zh, en, ja, etc.).
- Generate storyboard, character definitions, character reference sheet (PNG), page prompts, and page images.
- Partial workflows: storyboard only, prompts only, images only, regenerate specific pages.
- Edit, add, or delete pages with prompt file updates first.
- Backup existing prompt/image files before regeneration.
- Handle reference images: accept file paths or pasted images; ask for path if pasted without path; extract traits as text.
- Use
clarifytool for step 2 confirmations (one question at a time); handle timeouts by defaulting only that question and surfacing the default. - Detect and strip secrets (API keys, tokens, credentials) from source content.
Parameters
| Parameter | Meaning | Required | Default |
|---|---|---|---|
| Art | Art style / rendering technique | No | ligne-claire |
| Tone | Mood / atmosphere | No | neutral |
| Layout | Panel arrangement | No | standard |
| Aspect | Page aspect ratio | No | 3:4 portrait |
| Language | Output language | No | auto |
| Refs | File paths to reference images (used for style/palette trait extraction only, not passed to image model) | No | - |
| Storyboard only | Generate storyboard only, skip prompts and images | No | - |
| Prompts only | Generate storyboard + prompts, skip images | No | - |
| Images only | Generate images from existing prompts directory | No | - |
| Regenerate N | Regenerate specific page(s) only (e.g., '3' or '2,5,8') | No | - |
Full Comic Creation Workflow
The following checklist shows the complete workflow. Use it to track progress.
Comic Progress:
- [ ] Step 1: Setup & Analyze
- [ ] 1.1 Analyze content
- [ ] 1.2 Check existing directory
- [ ] Step 2: Confirmation - Style & options ⚠️ REQUIRED
- [ ] Step 3: Generate storyboard + characters
- [ ] Step 4: Review outline (conditional)
- [ ] Step 5: Generate prompts
- [ ] Step 6: Review prompts (conditional)
- [ ] Step 7: Generate images
- [ ] 7.1 Generate character sheet (if needed) → characters/characters.png
- [ ] 7.2 Generate pages (with character descriptions embedded in prompt)
- [ ] Step 8: Completion report
The high-level flow is:
Input → Analyze → [Check Existing?] → [Confirm: Style + Reviews] → Storyboard → [Review?] → Prompts → [Review?] → Images → Complete
Step 1: Setup and Analyze
Step 1.1: Analyze content — Produce analysis.md and source-{slug}.md. The slug is derived from the content topic. For example, a comic about Alan Turing gets slug alan-turing-bio.
Step 1.2: Check existing output directory — Check if comic/{topic-slug}/ already exists. If it does, append a timestamp to the slug to avoid conflicts. For example, if comic/alan-turing-bio/ exists, use turing-story-20260118-143052 as the slug.
Step 2: Confirmation (REQUIRED)
This step is required and must not be skipped. Use the clarify tool to ask the user one question at a time. Confirm the following:
- Art style (default:
ligne-claire) - Tone (default:
neutral) - Layout (default: standard)
- Aspect ratio (default: 3:4 portrait)
- Language (default: auto)
- Whether they want to review the outline (conditional step 4)
- Whether they want to review prompts (conditional step 6)
Timeout handling: If clarify times out on a question, default only that question and surface the default visibly in the next message. For example: "Style: defaulted to ohmsha preset (clarify timed out). Say the word to switch." Continue asking remaining questions sequentially. A timeout does not mean the user consented to all defaults.
Presets: Before recommending combinations, read the compatibility matrix and content-signal-to-preset table in references/auto-selection.md. When a preset is picked, load the preset's file from references/presets/. Presets have special rules beyond plain art+tone:
ohmsha: Usesmangaart +neutraltone; visual metaphors, no talking heads, gadget reveals.wuxia: Usesink-brushart +actiontone; Qi effects, combat visuals, atmospheric.shoujo: Usesmangaart +romantictone; decorative elements, eye details, romantic beats.concept-story: Usesmangaart +warmtone; visual symbol system, growth arc, dialogue+action balance.four-panel: Usesminimalistart +neutraltone + four-panel layout; follows 起承转合 structure; B&W with spot color; stick-figure characters.
Full art style, tone, and layout definitions are in references/art-styles/, references/tones/, and references/layouts/ respectively. Full workflow details in references/workflow.md; partial workflow options in references/partial-workflows.md. Core templates: analysis-framework.md, character-template.md, storyboard-template.md, ohmsha-guide.md.
Step 3: Generate Storyboard and Characters
Generate storyboard.md and character definitions in characters/characters.md. The storyboard outlines the page-by-page narrative. Character definitions describe each recurring character's appearance, personality, and role.
Step 4: Review Outline (Conditional)
Only perform this step if the user requested it in Step 2. Present the storyboard for user feedback and incorporate changes.
Step 5: Generate Page Prompts
Generate page prompts at prompts/NN-{cover|page}-[slug].md with character descriptions embedded inline from characters/characters.md. The prompt file naming convention is prompts/NN-{cover|page}-[slug].md where NN is a two-digit sequence number. The file pattern prompts/NN-{cover|page}-[slug].md is used for each prompt file, and the corresponding image file pattern is NN-{cover|page}-[slug].png. The directory prompts/ contains files matching prompts/…md, and the output images match …png.
Step 6: Review Prompts (Conditional)
Only perform this step if the user requested it in Step 2. Present the prompts for user feedback and incorporate changes.
Step 7: Generate Images
Step 7.1: Generate character reference sheet — Generate characters/characters.png if the comic is multi-page with recurring characters. Skip for simple presets (e.g., four-panel minimalist) or single-page comics. The prompt file must exist first. This PNG is a human-facing review artifact, not an input to image_generate.
Step 7.2: Generate page images — For each page, ensure the prompt file exists at prompts/NN-{cover|page}-[slug].md. Call image_generate with the prompt and aspect_ratio (mapped from the storyboard ratio). Download the returned URL to a local PNG using absolute paths for curl -o. The output file naming convention is NN-{cover|page}-[slug].png.
Critical: Always use absolute paths for curl -o when downloading images. Never rely on persistent-shell CWD across batches (silent footgun). The TERMINAL_LIFETIME_SECONDS environment variable may cause the working directory to change between batches. Using cd before curl -o relative/path.png can cause files to land in the wrong directory. Instead, specify the full path: curl -fsSL "" -o /abs/path/to/comic//NN-page-.png. Note that workdir= may produce paths like comic// (double slash), which is harmless.
Aspect ratio mapping: The tool supports these aspect ratios: 3:4, 9:16, 2:3, 4:3, 16:9, 3:2, 1:1. The aspect_ratio parameter passed to image_generate should be one of: portrait, landscape, square, or a specific ratio like 3:4.
Step 8: Completion Report
Provide a summary to the user listing the output directory, number of pages, and any notable decisions or defaults applied.
Page Modification
Edit a Page
Update the prompt file first, then regenerate the image, then download the new PNG. Backup existing files before regeneration.
Add a Page
Create the prompt at the desired position with character descriptions embedded. Generate the image. Renumber subsequent prompts and images. Update the storyboard.
Delete a Page
Remove the prompt and image files. Renumber subsequent files. Update the storyboard.
Reference Image Handling
Accepting References
Accept file paths or pasted images from the user. If file paths are provided, copy them to refs/NN-ref-{slug}.{ext} alongside the comic output. If a pasted image has no path, ask the user for the path via clarify, or fall back to verbal trait extraction.
Usage Modes
Record the usage mode (style, palette, scene) and extracted traits in each page prompt frontmatter (YAML). For example:
references:
- ref_id: 01
filename: 01-ref-scene.png
usage: style
traits: "muted earth tones, soft-edged ink wash, low-contrast backgrounds"
- For
styleusage: extract line treatment, texture, mood; append to every page prompt body. - For
paletteusage: extract hex colors; append to every page prompt body. - For
sceneusage: extract scene composition or subject notes; append to relevant page(s) only.
Important Constraint
The image_generate tool is prompt-only and does NOT accept reference images. Use reference images only to extract text traits. Character consistency is driven by text descriptions in characters/characters.md embedded inline in every page prompt (Step 5). The PNG character sheet (Step 7.1) is a human-facing review artifact, not an input to image_generate.
Failure Modes and Constraints
Image Generation
Image generation takes 10-30 seconds per page. Auto-retry once on failure.
Path Issues
Using relative paths for curl -o can cause files to land in the wrong directory if CWD changes between batches (silent footgun). Always use absolute paths. For example, instead of curl -o 06-page-skills.png, use curl -fsSL "" -o /abs/path/to/comic//NN-page-.png.
Timeout Handling
A clarify timeout does not mean the user consented to all defaults. Default only that one question and continue asking remaining questions. Surface the default visibly in the next message.
Backup Before Regeneration
Backup existing prompt and image files before regeneration. Use the suffix -backup-YYYYMMDD-HHMMSS. For example, rename prompts/01-page-intro.md to prompts/01-page-intro-backup-20260118-143052.md.
Prompt File Existence
The prompt file must exist before calling image_generate for any page or character sheet. Otherwise, reproducibility is lost.
Language
Use the user's input language for ALL interactions (storyboard outlines, prompts, user selections, progress updates, questions, errors, summaries). Technical terms remain in English. Detection priority: user-specified > conversation language > source content language.
Secrets Detection
Detect and strip secrets (API keys, tokens, credentials) from source content before writing any output file.
Sensitive Content
Use stylized alternatives for sensitive public figures.
Output Directory Structure
The output directory is comic/{topic-slug}/ and contains:
analysis.md— Content analysissource-{slug}.md— Source contentstoryboard.md— Storyboardcharacters/— Character definitions and reference sheetcharacters.md— Character definitionscharacters.png— Character reference sheet (PNG)
prompts/— Page promptsprompts/NN-{cover|page}-[slug].md— Individual page prompts
*.png— Generated page images (e.g.,NN-{cover|page}-[slug].png)refs/— Reference images (e.g.,refs/NN-ref-{slug}.{ext})
Examples
Example: Alan Turing Comic
User asks for a comic about Alan Turing. Slug: alan-turing-bio. Output directory: comic/alan-turing-bio/. If the directory already exists, append a timestamp: turing-story-20260118-143052.
Example: Reference Image for Style
User supplies a reference image for style extraction. The image is copied to refs/01-ref-{slug}.{ext}. Traits are recorded in page prompt frontmatter.
Example: Four-Panel Preset
User requests four-panel preset. Uses minimalist art + neutral tone + four-panel layout. Follows 起承转合 structure. B&W with spot color. Stick-figure characters.
Example: Ohmsha Preset
User requests ohmsha preset. Uses manga art + neutral tone. Visual metaphors, no talking heads, gadget reveals.
Example: Wuxia Preset
User requests wuxia preset. Uses ink-brush art + action tone. Qi effects, combat visuals, atmospheric.
Example: Shoujo Preset
User requests shoujo preset. Uses manga art + romantic tone. Decorative elements, eye details, romantic beats.
Example: Concept-Story Preset
User requests concept-story preset. Uses manga art + warm tone. Visual symbol system, growth arc, dialogue+action balance.
Example: Partial Workflow
User says "storyboard only". Skip steps 5-8 after step 3.
Example: Regeneration
User says "regenerate page 3". Backup existing prompt and image for page 3, update the prompt file, regenerate the image, download the new PNG.
Example: Timeout Handling
clarify times out on the first question. Default that one option, continue asking remaining questions, surface the default in the next message.
File Listing Commands
Use ls to list files in the output directory. Common patterns:
ls comic/{topic-slug}/— List all filesls comic/{topic-slug}/prompts/*.md— List prompt filesls comic/{topic-slug}/*.png— List generated images
Compatibility and Reference Files
Before recommending combinations in Step 2, read:
references/auto-selection.md— Compatibility matrix and content-signal-to-preset tablereferences/art-styles/— Full art style definitions (e.g.,references/art-styles/.md)references/tones/— Full tone definitions (e.g.,references/tones/.md)references/presets/— Preset files (e.g.,references/presets/.md)references/layouts/— Layout definitionsreferences/workflow.md— Full workflow detailsreferences/partial-workflows.md— Partial workflow options
Year 2026
The tool is designed to be used in the year 2026 and beyond, with timestamp-based conflict resolution using the format YYYYMMDD-HHMMSS.