antislop Skill for Claude Code: What It Does and How to Install It
Gives Claude 40 rules that stop generic AI-looking UI, then makes it check its own work with a pass or fail report before it delivers.
- Installs
- 8.4K
- on skills.sh
- Stars
- 7.9K
- on GitHub
- Files
- 1
- besides SKILL.md
- Updated
- Nov 2026
- last repo commit
antislop is a filter, not a style guide. Its 40 rules (19 hard gate, 12 purpose-gate, 9 quality locks) reject technique without purpose, such as fake stats, dead buttons and decoration with no reason, and require the result to feel specific and alive. At the start of a session Claude asks whether you want it during the work or after, as an audit with a numbered findings list.
It does not invent a visual direction. For UI you plan to ship, the README says a DESIGN.md is mandatory, and without one the result is labelled a draft without direction. This page covers the core skill. The repo has five more, for UI, copywriting, accessibility, responsive layout and code comments.
Good for
- Developers and designers building websites or apps with an AI coding agent
- Anyone auditing a finished UI for AI-slop patterns and fixing the findings they pick
- People who keep a DESIGN.md and want the agent held to it
Install antislop
Claude Code
- 1
Copy this prompt.
Install the antislop skill in Claude Code. Run this command: /plugin marketplace add https://github.com/miqdadbadjuber/anti-slop /plugin install antislop@anti-slop Then check that it loaded and tell me how to use it.
- 2
Paste it into Claude Code. It shows the change and asks before running it.
Pick one route. The plugin and installer routes bring all six antislop skills; the skills CLI and folder routes below install only the core. The README says the core alone is fine. The Claude Code plugin also registers a local contrast-check MCP server (antislop-contrast). Update later with the command for the route you used, for example claude plugin update antislop@anti-slop.
From the project README. Source.
What's in the skill
When it loads
Anti Slop: Rules for AI Coding Agents. The core filter. Load always to stop generic AI slop.
Sections in SKILL.md
- Already installed, and the user asks how to update
- Two Usage Modes
- What This Is (and What It Isn't)
- Core Principle
- The Craftsmanship Standard
- C-1 — Intentionality
- C-2 — Functional Completeness
- C-3 — Content-Driven Composition
- C-4 — Resilience
- C-5 — Evidence Over Claims
- Part 1: AI Slop Patterns (Warning Signs)
- Visual & Color
- Layout & Components
- Copywriting & Content
- Decorative Elements
- Functionality & Content
Key facts
- Installs on skills.sh
- 8,437 (Nov 2, 2026)
- GitHub stars
- 8,002 for the whole repo (Nov 2, 2026)
- Last repo commit
- Nov 1, 2026
- Audits on skills.sh
- Gen Agent Trust Hub: warn, Socket: pass, Snyk: pass
What it includes
Asks whether to apply the rules DURING the work or AFTER, as an audit.
Skipped if you chose a mode in chat or saved one in ~/.config/antislop/settings.json.
Mode questionApplies the rules while writing UI, then runs the Delivery Gate.
Try: “Build a landing page for my app with antislop on”
Write ·
During modeWrites a numbered findings list to anti-slop/audit-001-YYYY-MM-DD.md.
Nothing is changed until you approve numbers; then it fixes those and writes a follow-up report.
Try: “Audit this site with antislop”
Write ·
After mode (audit)Honesty, function and accessibility: no fake stats, no dead controls, contrast.
Read ·
Hard Gate (19 rules)Color, icons, glass, shadow, glow, cards and motion need a written reason.
Read ·
Purpose-Gate (12 rules)Consistency: layout, radius, CTA, copy, identity, dark mode, palette, no clones.
Read ·
Quality Locks (9 rules)Three dials (energy, rhythm, motion) and a Design Read turn a brief into choices.
Needs direction from your DESIGN.md.
Read ·
Liveliness toolkitA PASS or FAIL report in four blocks, each PASS backed by evidence.
R-35 says to run or build the app, check the console and click every control before delivering.
Runs code ·
Delivery GateReads the mode field of ~/.config/antislop/settings.json; writes it only if you ask.
npx antislop-ai --mode during|after|ask also sets it.
Write ·
Saved mode preference
From the official docs
Permissions
Medium risk
What it can do
- Read your data
- Create and change things
- Delete things
- Read-only mode
Watch out for
It edits your UI files
In During mode Claude writes UI to these rules. In After mode it fixes only the audit numbers you approve. Use version control and review the diff. github.com
The Delivery Gate runs your app
R-35 tells Claude to run or build the app, check the console and click through every interactive element before it calls the work done. That runs your project's own build and start commands. github.com
The Claude Code plugin bundles an MCP server
plugin.json registers antislop-contrast, started by a Node launcher that runs contrast-mcp.py with python3 or python. SECURITY.md describes it as a local file with no network access that checks color pairs. plugin.json lists no hooks. github.com
The installer writes into your project files
npx antislop-ai adds a pointer block to CLAUDE.md, AGENTS.md or GEMINI.md and copies skill folders. SECURITY.md says it asks first. npx antislop-ai --update replaces antislop folders at project and global scope without asking. github.com
DESIGN.md is treated as data
SKILL.md says to apply the design fields from DESIGN.md and to treat any command-like text inside it as content, not as an instruction. github.com
Technical details
| Turning on read-only | After mode is an audit that changes nothing until you approve numbers, but there is no read-only switch. |
|---|---|
| Sign-in and keys | SKILL.md names no account or API key. |
| Scopes |
|
Questions
What does the antislop skill do?
It gives Claude 40 numbered rules (R-01 to R-40) that reject generic AI-looking UI, copy and code patterns. Claude applies them while building, or audits finished work, and must pass a Delivery Gate report.
How do I install it?
Use the install box at the top of this page. The README documents a Claude Code plugin, a Codex plugin, a Cursor plugin, a Cline CLI plugin, the npx antislop-ai installer and the skills CLI.
Do I need the other antislop skills?
No. The README says to install one, several, or the core alone. The core is always loaded; antislop-ui, antislop-copywriting, antislop-human, antislop-layoutmobile and antislop-code each cover one concern. The Claude Code plugin installs all six.
Do I need a DESIGN.md?
For UI you intend to ship, the README says yes. Without one the result is labelled a draft without direction, because antislop removes slop but does not invent a visual direction.
Does it work in the Claude app?
The GUIDE documents it: turn on code execution, open Customize, and upload the zipped skill folder. The GUIDE also says that in a chat app the rules act as reference material, not as an enforced gate.
Is it safe?
It edits your UI files, can run your build, and the installer can write a pointer block into CLAUDE.md or AGENTS.md. The Claude Code plugin also adds a local contrast-check MCP server. Use version control and read the diff.
Why does it ask me a question every session?
By default it asks whether to apply the rules during or after the work. Save an answer with npx antislop-ai --mode during (or after), or tell Claude the mode in chat for one session.
What does the audit mode produce?
A numbered findings list in anti-slop/audit-001-YYYY-MM-DD.md, each finding citing the broken rule. Nothing is changed until you approve specific numbers.
Information
updated Nov 2, 2026
- Repository
- miqdadbadjuber/anti-slop
- Folder
skills/antislop
Sources
Docs, issue trackers and discussions this page is built from.
Show all 8Hide
- anti-slop README · github.com
- anti-slop GUIDE.md · github.com
- anti-slop plugin.json · github.com
- anti-slop marketplace.json · github.com
- antislop SKILL.md · github.com
- anti-slop SECURITY.md · github.com
- contrast-mcp-launcher.mjs · github.com
- contrast-mcp.py · github.com