Character Image Generator (character-image-generator)

This skill is optimized for character identity, silhouette clarity, costume logic, and reusable character prompt structure.

This skill keeps the same single-gateway runtime, readiness gate, model-selection flow, and CLI behavior as image-generation, but narrows the briefing and prompt construction for character image generator work.

Safety & Scope

• Network: This skill calls the WeryAI gateway over HTTPS (https://api.weryai.com).
• Auth: Uses IMAGE_GEN_API_KEY. The key is never printed. It may be persisted only when you explicitly run npm run setup -- --persist-api-key.
• Reference images: Must be public URLs (https:// recommended). http:// may work but is insecure. Local file paths and data: URLs are rejected.
• No arbitrary shell: The generation runtime does not execute arbitrary shell commands.
• Files written: Output images and optional local config under .image-skills/character-image-generator/ (project) and/or ~/.image-skills/character-image-generator/ (home).

Use Cases

• original character portraits
• fantasy or sci-fi hero concepts
• character sheets and reference images
• mascot or spokesperson characters
• narrative cast exploration

First Trigger Rules

Before the first generation run in a new project or environment:

1. Run npm run ensure-ready -- --project . --workflow <workflow>
2. If runtime dependencies are missing, ask for approval and install them
3. If IMAGE_GEN_API_KEY is missing, offer to configure it now
4. If no model is configured yet, initialize Nano Banana 2 (GEMINI_3_1_FLASH_IMAGE) as the default

Do not ask the user to edit config files manually. Treat API keys as secrets and never echo them back.

Clarify These Decisions

Ask one question at a time. Prioritize:

1. character role or archetype
2. age range, gender expression, and body language
3. setting: fantasy, modern, sci-fi, historical, etc.
4. signature outfit, props, and color identity
5. portrait, half-body, full-body, or sheet layout

Recommended Defaults

• aspect ratio: 3:4 or 4:5
• recommended style: anime, editorial, manga, or photoreal
• composition: one primary character, readable silhouette
• background: simple enough that outfit and pose stay dominant

Prompt Blueprint

Build the prompt in this order:

1. Identity: role, archetype, age range, personality, faction, or profession.
2. Look: face, hair, body language, posture, expression.
3. Outfit + props: signature materials, colors, weapon/tool, emblem, etc.
4. Presentation: portrait, full body, turnaround, or concept sheet.
5. World feel: fantasy, cyberpunk, school life, mythology, noir, etc.

Use one clean prompt direction at a time instead of mixing many competing ideas.

Prompt Rules

• Always define silhouette, costume anchors, and one memorable visual hook.
• If the user wants a reusable character, keep accessories and palette consistent across variants.
• Use a sheet-style prompt for reference art; use a scene-style prompt for story art.
• When underspecified, prefer strong archetypes over generic 'cool character' wording.

Avoid

• too many unrelated props
• multiple equal-priority characters unless the user asks for a duo/group
• unclear costume materials or indistinct silhouette
• overly busy background scenes for reference-style art

Workflow

1. Run the readiness gate and resolve IMAGE_GEN_API_KEY
2. Clarify the scenario-specific decisions above
3. Build a single strong prompt from the blueprint
4. Choose a recommended style only if it helps the request
5. Generate the image
6. If the user wants variations, change one major variable at a time and re-generate

Script

{baseDir} is the directory containing this file. ${BUN_X} is either bun or npx -y bun.

Usage Examples

# examples only; M should be chosen by the user or resolved by the agent
M=<chosen model key>

${BUN_X} {baseDir}/scripts/main.ts --prompt "original fantasy character, young female archivist mage, emerald robe with gold embroidery, floating paper charms, calm intelligent expression, full-body concept art, clean parchment backdrop" --style editorial --image character.png --ar 3:4 -m "$M"

${BUN_X} {baseDir}/scripts/main.ts --prompt "anime character sheet for a cyberpunk courier boy, silver undercut hair, orange visor, black utility jacket, messenger drone companion, front pose with detail callouts feel, highly consistent design" --style anime --image character-sheet.png --ar 4:5 -m "$M"

Delivery Rules

• Tell the user what you are generating and which model is being used before you start
• Show the image directly when it is ready; do not reply with only a filename
• If the user asks for revisions, only change the necessary direction instead of restarting everything
• If the request is underspecified, use the clarification order above before writing the final prompt

References

• references/config/first-time-setup.md
• references/config/preferences-schema.md
• references/config/model-registry-schema.md
• references/style-presets.md
• references/weryai-platform.md