| name | chapter-illustration-prompter |
| description | Generates chapter illustration prompt files (assets/chapters/chapter-XX.illustration-prompt.md). Activates when the user asks to create illustration prompts, scene art guides, or visual references for a chapter. Produces a Master Chapter Prompt, 3–6 scene-specific prompts, and technical notes. |
Chapter Illustration Prompter Skill
TRIGGER: Whenever the user asks to create illustration prompts, scene art, or visual guides for a chapter.
Instructions for AI
You generate assets/chapters/chapter-XX.illustration-prompt.md files. Before generating, you MUST:
- Read the chapter — know every scene, character, and setting that appears.
- Read character design sheets — check
assets/characters/ for every character in the chapter. Pull their skin tone, clothing, and design notes.
- Read the constitution — confirm the visual tone (dead serious, epic mundane, Solo Leveling aesthetic).
- Scan for knowledge sources — check all of the following locations for any relevant reference material and incorporate what you find into the prompts:
| Location | What to look for |
|---|
memory/constitution.md | Core visual tone, style rules, non-negotiables |
stories/[novel]/specification.md | Character descriptions, world details, setting facts |
stories/[novel]/creative-plan.md | Chapter intent, mood targets, foreshadowing goals |
assets/characters/ | Character design sheets (skin tone, clothing, build) |
.gemini/skills/genre-knowledge/ | Genre-specific visual conventions (manhwa, anime, etc.) |
knowledge/ | Any project-specific visual references, world-building notes, lore docs |
knowledge-base/ | Same — alternative folder name |
references/ | Same — alternative folder name |
docs/ | Same — scan for any visual or lore reference files |
memory/ | Any additional reference files beyond constitution |
If a folder does not exist, skip it silently. If any relevant file is found, read it and extract any detail that improves visual accuracy — character descriptions, location layouts, color palettes, cultural details, technology descriptions, or tone references. Do NOT ignore these sources even if the user did not explicitly mention them.
Missing Design Sheet Protocol
For every character in the chapter, check if a corresponding design sheet exists in assets/characters/[slug]-design-sheet.md.
If a design sheet is NOT found:
- Notify the user immediately before generating:
⚠️ No design sheet found for [Character Name]. Guessing appearance from chapter context. For better consistency, create a design sheet at assets/characters/[slug]-design-sheet.md.
- Proceed anyway — infer their appearance from all available chapter text, their rank, role, and any description in the specification.
- Make the guessed appearance explicit in the prompt (do not leave it vague).
- List all characters missing design sheets at the top of the output file as a warning block.
Output File Structure
Every illustration prompt file MUST contain these sections in order:
1. YAML frontmatter
2. Illustration Overview table
3. Master Chapter Prompt
4. Individual Scene Prompts (3–6)
5. Technical Notes
6. Placement Format
1. Frontmatter
---
name: chapter-XX-illustration-guide
description: "Visual illustration guide for Chapter XX of Solo Working."
---
2. Illustration Overview Table
Summarize all scenes before writing prompts. Anchor each to a direct quote from the chapter text.
| # | Scene | Character(s) | Placement | Purpose |
| --- | ------------ | ---------------- | -------------------------- | ------------------------- |
| 1 | [Scene name] | [Character name] | At "[exact chapter quote]" | [One-line visual purpose] |
3. Master Chapter Prompt
One cinematic cover-style image summarizing the chapter's emotional arc.
Critical rule: The Master Chapter Prompt must be a fully self-contained image generation prompt. An AI with no other context — no design sheets, no chapter text — must be able to generate the correct image from the Master Prompt alone. It is NOT a brief summary. It is the most important single prompt in the file and must be the most detailed.
Apply ALL Detail Depth Rules (see Section 4) to every element in the Master Prompt — the same exhaustive standard used for individual scene prompts. Never leave characters, objects, or environment vague.
Structure:
- Foreground: Primary character(s) fully described per Detail Depth Rules — exact skin tone, exact clothing (fabric, color, condition), hair color/length/style, eye color, pose, expression, any props. Include Rank label and any floating System UI elements.
- Midground: Secondary characters or elements — full Detail Depth description for each character present. Spatial relationship to foreground. Any System windows, overlays, or environmental phenomena.
- Background: Full environment breakdown — room type, dimensions, wall/floor/ceiling materials and colors, every light source (type, position, color temperature, shadow direction), furniture and fixtures, atmospheric effects (glow, dust, heat distortion), depth layers.
- Mood & Style: 2–3 sentences on emotional tone and visual atmosphere. Reference Solo Leveling manhwa aesthetic. State dominant color clash.
- Composition: Camera angle (low-angle / eye-level / high-angle), framing (tight / medium / wide), focal hierarchy (what draws the eye first, second, third).
- End with:
**No text. No watermarks. No logos.**
4. Individual Scene Prompts
Repeat for each scene in the overview table. Each prompt has exactly 5 sections:
- Location in Chapter: Exact quote trigger from the chapter text.
- Scene Description: 2–4 sentences — what is happening, who is present, what is visible.
- Visual Elements: Exhaustive bullet list — see Detail Depth Rules below.
- Mood: 1–3 sentences — visual emotion, not plot summary.
- Composition: Camera angle, framing, focal hierarchy.
Detail Depth Rules (MANDATORY)
Every Visual Elements section MUST describe objects, characters, and environment at full resolution. Never leave anything vague. Apply these rules:
For every CHARACTER present:
- Full name + Rank label
- Exact skin tone (from design sheet or inferred)
- Exact clothing — fabric, color, condition, fit, any wear or damage
- Hair color, length, style
- Eye color and expression
- Pose — what are they doing with their hands, body weight, facing direction
- Emotional state visible on face and body
- Any props they are holding or wearing
For every OBJECT in the scene:
- Material (metal, paper, glass, fabric, plastic)
- Color and finish (matte, glossy, worn, new)
- Size relative to the character
- Condition (pristine, worn, damaged, aged)
- Position in the frame (foreground, mid, background; left, center, right)
- Any text or markings visible on the object (if relevant to the scene)
For the ENVIRONMENT:
- Room type and approximate dimensions (cramped, spacious, corridor-narrow)
- Wall, floor, ceiling materials and colors
- Every light source — type, position, color temperature, direction of cast shadows
- Furniture and fixtures — describe each present item briefly
- Atmospheric effects — dust, steam, heat distortion, digital glow, smoke
- Depth layers — what is near, middle, far
Negative space: Note what is deliberately empty or dark to create contrast.
5. Technical Notes
Include at the end of every file:
- Color Palette: Primary neutrals, secondary tones, accent/glow colors. State warm/cold rule for this chapter.
- Lighting Consistency: Map each light source to its color temperature (fluorescent → harsh white, monitor → blue-green, System UI → neon blue, city glow → warm amber).
- Character Design References: List every character in the chapter. For each: link to design sheet if it exists, or write a full inferred appearance description. Mark missing sheets with
⚠️ No design sheet — appearance inferred.
- Artistic Style: Cinematic manhwa, high contrast, anime character design, Solo Leveling visual language, no caricature.
6. Placement Format
Fixed block — copy exactly into every file:
[ILLUSTRATION #X: {Title}]
_{One-line caption}_
Scene Selection Rules
Prefer these scene types:
- Character introduction (first appearance)
- Spatial anchor (first time a key location appears)
- Power dynamic (hierarchy/rank made visible)
- Turning point (catalyst or revelation)
- Internal moment (protagonist thinking/calculating)
- System event (UI appearance, awakening, quest notification)
- Confrontation (direct conflict between characters)
Avoid:
- Generic dialogue with no visual hooks
- Scenes already captured by the Master Prompt
- More than 6 illustrations per chapter
Character Design Rules
These are mandatory for every prompt:
- Never write "Filipino facial features." Always use "anime character design aesthetic" or "anime protagonist aesthetic."
- Always include the character's Rank next to their name:
Juan dela Cruz (Rank-F).
- Always state skin tone — pull from
assets/characters/[slug]-design-sheet.md.
- Background characters default to:
anime-style proportions, face ambiguous or silhouetted.
- System UI elements render as:
semi-transparent neon-blue, geometric, clean-edged, no decorative borders.
- Always end Master Prompt with:
**No text. No watermarks. No logos.**
Mood Vocabulary
Use this language in Mood sections to maintain the Solo Working tone:
- Rising: quiet competence, coiled readiness, contained intensity, predatory focus, invisible preparation
- Falling: institutional indifference, crushing hierarchy, the silence of being overlooked, bureaucratic inevitability
- Turning: the moment before catastrophe, the pivot point, foreshadowing made visual, before everything changes
- System: the mundane rendered epic, the dungeon in the office, awakening rendered clinical, power discovered in fluorescent light
Composition Quick Reference
| Type | When to Use |
|---|
| Wide establishing | First appearance of a space |
| Low angle hero | Protagonist empowerment |
| Over-shoulder | Power dynamic, two-character scenes |
| Close-up | Internal moments, System UI |
| Top-down | Planning, calculation, overview |
| Split-light | Duality, tension between two worlds |
| Silhouette | Background/mystery characters |
Global Style Block (for External AI Tools)
When prompting an external AI (Midjourney, DALL-E, Stable Diffusion, Gemini Imagen) without project context, prepend:
Visual style for "Solo Working" (web novel series):
- Aesthetic: Manhwa/anime hybrid. Solo Leveling visual language.
- Characters: Anime character design. Expressive eyes. Stylized proportions. High-contrast shading.
- Setting: Modern Philippine corporate office (BGC/Makati). Rendered with epic dungeon weight.
- Tone: Dead serious. No comedy. Mundane details treated as life-or-death.
- Lighting: Cold fluorescent dominant. Neon-blue for System events. Warm amber only from city lights.
- Color: Dark gray, charcoal, deep navy base. Neon-blue/red accents. High contrast.
- NO text, watermarks, logos, or speech bubbles.
Output Checklist
Before finalizing, verify:
- Frontmatter YAML present and correct
- Missing design sheet warnings shown to user before generation begins
- Overview table anchors every illustration to a chapter quote
- Master Chapter Prompt covers the full emotional arc
- All scene prompts have the 5 required sections
- Every Visual Elements section fully describes all characters, objects, and environment — nothing is vague or generic
- Every character in a scene has: name + Rank, skin tone, exact clothing, hair, eyes, pose, expression
- Every object has: material, color, condition, size, position in frame
- Technical Notes include chapter-specific color/light additions
- Character Design References list every character — with design sheet link or full inferred description + warning
- Placement Format block present
- No "Filipino facial features" language anywhere
- All Rank labels present in character descriptions
- Master Prompt ends with no-text declaration