miqdadbadjuber/anti-slop

What it solves

It prevents AI coding agents from generating "AI slop"—generic, sterile, and predictable UI designs and copy. Instead of providing a rigid style guide, it acts as a filter that forces agents to justify their design choices and maintain a sense of "liveliness" in their output, ensuring the results feel human and specific rather than generic.

How it works

The project provides a set of mandatory rules (divided into Hard Gates, Purpose-Gates, and Quality Locks) and a "Liveliness Toolkit" that agents must follow. It is implemented as a collection of "skills" (Markdown files following an open standard) that can be installed into AI agents like Claude Code, Codex, or Antigravity.

Users can operate in two modes:

• During: The rules guide the agent while it is building new UI.
• After: The agent audits existing work and provides a findings list for the user to approve and fix.

Who it’s for

Developers and designers using AI coding agents who want to avoid the generic look and feel of AI-generated interfaces and text.

Highlights

• Filter, not a style guide: Does not prescribe specific colors or fonts, but rejects technique without purpose.
• Tiered Rule System: Uses 38 mandatory rules to gate quality and consistency.
• Delivery Gate: Requires a mandatory PASS/FAIL report before any work is shipped.
• Modular Skills: Specialized rule sets for UI, copywriting, human-centric design, and mobile layout that can be loaded as needed.