A sophisticated framework that equips AI agents with the creative judgment and design taste necessary to produce and evaluate professional-grade visual content.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install creative-eye
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install creative-eye using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Creative Eye?
Creative Eye is a technical framework designed to solve the problem of mediocre AI-generated content. Most agents fail at design because they lack benchmarks and the ability to self-evaluate. By implementing this skill, agents move beyond simple generation and adopt the role of a creative director, using structured logic to analyze, compare, and refine visual assets.
Built on the STUDY → COMPARE → CREATE → EVALUATE sequence, this integration utilizes vision-capable models to perform objective audits against established brand standards. By leveraging these Openclaw Skills, developers can ensure that their agents maintain brand consistency and high aesthetic quality across merch, social posts, and product photography without constant human intervention.
Creative Eye Use Cases
- Automating design reviews for social media graphics and digital advertisements.
- Establishing creative quality gates to prevent amateur or off-brand content from being published.
- Building and maintaining a structured reference library and JSON-based style guides.
- Training AI agents to develop aesthetic judgment through daily design study logs.
- Refining AI-generated merchandise designs through iterative vision-model feedback loops.
How Creative Eye Works
- Study Phase: The agent analyzes high-tier brand examples to build a vocabulary of design principles and pattern libraries.
- Reference Benchmarking: Before generating content, the agent identifies specific visual references to serve as creative constraints.
- Intentional Creation: Visual assets are generated with specific rationales for every font, color, and layout choice.
- Automated Evaluation: Outputs are processed through a 5-point scorecard and 10-point pre-publish checklist using vision models.
- Self-Refine Loop: The agent performs up to three iterations of critique and adjustment to reach professional quality standards.
Creative Eye Setup
To deploy Creative Eye within your environment, initialize the reference directory structure and create the essential configuration files.
mkdir -p reference-library/{merch,social-content,product-photography,brand-systems,style-profiles,failures}
touch brand-guardrails.md style-guide.md creative-log.md
Ensure your agent is connected to a vision-capable LLM and has read/write access to these directories to store its evolving Openclaw Skills style profiles and study logs.
Creative Eye Data Schema & Taxonomy
Creative Eye organizes its aesthetic intelligence through a structured metadata taxonomy and filesystem.
| Data Component | Purpose | Format |
|---|---|---|
| Style Profiles | Captures aesthetic DNA (colors, type, lighting) from references. | JSON |
| Brand Guardrails | Defines hard and soft stops for automated quality gates. | Markdown |
| Reference Library | Categorized benchmarks including 'anti-examples' for what to avoid. | Directory |
| Study Log | Ongoing repository of design observations and failure analyses. | Markdown |
JSON profiles specifically track keys such as color_palette, typography, and composition to ensure generation prompts remain grounded in brand reality.
Creative Eye Advanced Features
- JSON Style Profile Extraction: Automatically translates visual references into structured data for high-fidelity prompt engineering.
- Self-Refine Vision Loop: Implements the Reflection pattern to iteratively improve visual outputs based on model-led critiques.
- Brand Guardian Pattern: Sets up automated quality gates that trigger blockages for logo violations or color drift.
- Root Cause Analysis (RCA): A workflow for documenting creative failures to prevent the agent from repeating the same aesthetic mistakes.
- Aesthetic DNA Analysis: Uses vision models to deconstruct successful designs into actionable metadata for future Openclaw Skills generation tasks.
SKILL.md
Loading
Related Openclaw Skills
Agent Self-Introduction Skill
A narrative identity skill that enables AI agents to introduce themselves clearly, consistently, and contextually.
Tilt Protocol
Tilt Protocol allows AI agents to manage self-custodied investment vaults holding tokenized synthetic stocks on the Robinhood L2 network.
SIGNL4 Alert Management
A specialized skill designed to automate the sending and resolving of SIGNL4 alerts via inbound webhooks to streamline incident response.
Agent Skills Security Tools
A specialized security auditing suite designed to scan, validate, and secure AI agent skill packages against vulnerabilities and credential leaks.
Didit Age Estimation
A facial analysis skill that estimates age and performs liveness detection to verify user identity and age compliance.
Didit AML Screening
A real-time compliance tool for screening individuals and companies against 1,300+ global watchlists and risk databases.
Featured*
