AntiSlop.
Deterministic rules and liveliness filters for AI coding assistants. Stops coding agents from generating generic "AI slop" UI and robotic copy, without letting the result turn sterile.
Overview & Philosophy
AntiSlop is a filter, not a style guide: it does not prescribe colors, fonts, or rigid layouts. Instead, it systematically strips away the mathematical average of the Internet that LLMs gravitate toward (purple gradients, rounded pill badges, low contrast text, and robotic corporate copy) while allowing your project's custom design rules to provide the creative soul.
Core Architecture & Tiers
The engine enforces 38 mandatory rules categorized into three strict operational tiers:
- Hard Gate: Absolute prohibitions (no generic purple gradients, no em dashes, no robotic markdown tags).
- Purpose-Gate: Techniques allowed only when explicit functional rationale is provided.
- Quality Locks: Strict cross-file consistency for typography, border contrast, and interactive feedback.
- Delivery Gate: Mandatory PASS or FAIL verification report executed before any code ships.
Modular Agent Skills
| Skill Name | Coverage & Scope |
|---|---|
| antislop | Core filter, rule tiers, Delivery Gate, and liveliness dials |
| antislop-ui | Visual layout, contrast ratios, components, motion, structure |
|
antislop- copywriting |
Headlines, CTAs, tone, eliminating fake stats and AI clichés |
| antislop-human | Accessibility, contrast checker, keyboard focus, interactive states |
|
antislop- layoutmobile |
Responsive breakpoints, touch targets, grids, zero horizontal overflow |
| antislop-code | Removes robotic AI code comments while preserving genuine documentation |
How to Install & Run
antislop is built upon the open Agent Skills standard and supports Claude Code, Antigravity CLI, Codex, and Cursor:
Operational Modes
antislop operates in two dedicated modes chosen at the start of any session:
Guides generation actively as files are being written, concluding with a strict Delivery Gate pass/fail validation.
Audits existing codebases, producing a prioritized findings report with actionable refactor recommendations.