Anti-Slop Open-Source Project Aims to Filter Generic AI Coding Outputs

Overview of Anti-Slop

A publicly available project titled anti-slop, hosted on GitHub by user miqdadbadjuber, has emerged to help filter out generic, AI-shaped UI designs, text, and code produced by AI coding agents. According to the project documentation, the utility functions as a filter rather than a rigid style guide, meaning it does not prescribe specific colors, fonts, or layouts, leaving stylistic direction entirely to the user.

Core Rules and Structure

The repository outlines 40 mandatory rules categorized into three tiers:

• Hard Gate
• Purpose-Gate
• Quality Locks

Additionally, it features a "Liveliness Toolkit" comprising three dials—Energy, Rhythm, and Motion—alongside a Design Read to encourage more dynamic outputs. Before any code is shipped, a mandatory Delivery Gate provides a pass/fail report split into four blocks.

Operation Modes and Agent Compatibility

Anti-slop can operate in two distinct modes during an agent session:

1. During Mode: Guides the work as it is being built, culminating in a Delivery Gate report.
2. After Mode: Audits existing work through a numbered findings list for user review and remediation.

The project integrates with various AI coding assistants and CLI tools, including Claude Code, Cursor, GitHub Copilot, and Gemini CLI, among others. Installation can be handled via automated setup scripts, manual file pasting, or through community directories like skills.sh.

Important Caveats and Verification

The documentation stresses that anti-slop does not supply creative direction; a separate `DESIGN.md` file is required for UI projects to avoid sterile results. Furthermore, the project maintainers explicitly note that the tool has no associated cryptocurrency, tokens, NFTs, or fundraising campaigns, warning that any financial solicitations using the name are unassociated scams.

To verify original claims and review installation instructions, interested users can consult the official GitHub repository at `miqdadbadjuber/anti-slop` or the project website.