AgentSkillsCN

comfyui-image-generator

新项目资产（人像、图标、竞技场美术）的设计与生成流程。当实施过程中需要一种尚未存在于 `assets/` 中的视觉资源时，可使用此流程。

SKILL.md
--- frontmatter
name: comfyui-image-generator
description: Design and generation workflow for new project assets (portraits, icons, arena art). Use when the implementation requires a visual resource that does not yet exist in `assets/`.

ComfyUI Image Generator

Engineer high-fidelity art assets consistent with the "Brutal" game aesthetic.

Workflow

  1. •Style Lookup: Inspect current textures in assets/textures/ to align lighting and grit.
  2. •Prompt Engineering: Use the prompts.md patterns to construct species-aligned tokens.
  3. •Workflow Selection:
    • •Load text_to_image_standard.json.
    • •Configure seed and target aspect ratio.
  4. •Validation: Check output against naming conventions (e.g., por_orc_warrior.png).

Guidelines

  • •Consistency: Use the same base model (SDXL) for all core species.
  • •Batching: Generate 4-8 variations before finalizing a choice.

Artefacts to Update (new assets)

  • •When adding new assets, place them under assets/textures/ and update any .tres resources that reference them.
  • •Add a line to ./docs/todo/master_todo.md indicating the new assets and which scenes or resources will use them.
  • •Commit assets in a separate commit and include small preview images in the PR description for faster review.