A bouncer that stops AI agents from shipping slop

What it does

Anti-slop is a set of 38 mandatory rules packaged as agent skills. It sits between an AI coding agent and its output, blocking generic “AI slop” patterns in UI, text, and code comments. The core filter is augmented by optional skills for specific concerns like layout, copywriting, accessibility, and mobile design. Before anything ships, a mandatory PASS/FAIL Delivery Gate forces the agent to prove the work is alive and intentional, not merely clean.

The interesting bit

Instead of prescribing colors, fonts, or layouts, it uses a “Liveliness Toolkit” with three dials—ENERGY, RHYTHM, and MOTION—and requires your own DESIGN.md to supply actual direction. It treats sterility as a missing-direction problem, not a filter failure. That distinction keeps it from becoming the very kind of rigid template it is trying to prevent.

Key highlights

• 38 rules across three enforcement tiers: Hard Gate, Purpose-Gate, and Quality Locks.
• Additive skill system: load only the concerns the task needs (UI, copy, code comments, mobile, human/accessibility).
• Two usage modes: During guides work as it is built; After audits finished output with a numbered findings list.
• Mandatory Delivery Gate report in four blocks, including a click-through smoke test for interactive elements (R-35).
• Drops into seven agents via the open Agent Skills standard or a single antislop.md file for manual use.

Verdict

Worth a look if you are tired of AI-generated bento grids, fake terminal windows, and CTAs that read like a chatbot wrote them. Skip it if you want an AI to hand you a complete visual identity; this filter demands you bring your own taste.