IP as Logo: Consistent Mascot Skill Released

IP as Logo is an agent skill designed to solve the consistency problem that arises when creating corporate mascots with image generation models. Rather than relying on better models, it compresses guidelines that consistently enforce mascot logo constraints for generative models into a single skill.

Enforced Design Rules

Instead of instructing the generative model on what to draw, the skill applies a list of rules defining what to prohibit.

• Single Silhouette: Maintain thick, rounded shapes composed of 4–7 large basic geometric forms.
• Meaningful 3 Colors: Limit to two colors for the IP body and one background color.
• Corner Entry Composition: Enter from the bottom-left or bottom-right corner, filling 85–95% of the square.
• Friendly Animals First: Prioritize animal mascots that have broad appeal over abstract objects like clocks or locks.

Generation Workflow and Installation

It runs within a product repository and, if context is sufficient, presents 3 directions and proposes 6 images generated independently. The first candidate for each direction is placed to enter from the bottom-left, and the second from the bottom-right, ensuring balance. Generation is treated as a one-time creative lottery, and the skill does not automatically retry results.

It can be installed via the command npx skills@latest add s1dashu/ip-as-logo-skill and is usable in agents with built-in image generation capabilities, such as Codex, Coze, Doubao, YouMind, Manus, Gemini Apps, and Replit Agent. It is released under the MIT license and is free for personal and commercial use.