Back to search
slop-skill
MarketplaceWorks with Claude CodeGitHub
Official marketplace catalog for slop-skill — calibrated AI slop engineering for web interfaces.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Checked today
Badge for your README
Markdown
[](/listings/seohuda--slop-skill)HTML
<a href="/listings/seohuda--slop-skill"><img src="/api/listings/seohuda--slop-skill/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
Agent Skills CLI (`npx skills`)
# Add all skills from the repository npx skills add seohuda/slop-skill # Target slop-skill explicitly or install globally npx skills add seohuda/slop-skill --skill slop-skill npx skills add seohuda/slop-skill --global
Claude Code Plugin Marketplace
Inside Claude Code:
/plugin marketplace add seohuda/slop-skill /plugin install slop-skill@slop-skill
Local Symlink Install (`install.sh`)
git clone https://github.com/seohuda/slop-skill.git cd slop-skill ./install.sh
By default, `install.sh` symlinks `skills/slop-skill` into `${CLAUDE_SKILLS_DIR:-$HOME/.claude/skills}/slop-skill`. It refuses to overwrite existing files or conflicting directories unless `--force` is passed.
Useful flags:
./install.sh --dry-run # Preview actions without modifying disk ./install.sh --target /custom/skills # Install into a custom directory ./install.sh --force # Replace an existing path at the target
---
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 19 lines
- {
- "$schema": "https://json.schemastore.org/claude-code-marketplace.json",
- "name": "slop-skill",
- "version": "0.1.0",
- "description": "Official marketplace catalog for slop-skill — calibrated AI slop engineering for web interfaces.",
- "owner": {
- "name": "seohuda",
- "url": "https://github.com/seohuda"
- },
- "plugins": [
- {
- "name": "slop-skill",
- "description": "Intentional AI slop design system with 5-level slop intensity calibration, deterministic generator CLI, and 30+ rule Slop Meter.",
- "version": "0.1.0",
- "source": "./",
- "category": "design"
- }
- ]
- }
README.md · README · 265 lines
- # slop-skill
- <p>
- <a href="README.ko.md"><img src="https://img.shields.io/badge/README-Korean%20(%ED%95%9C%EA%B5%AD%EC%96%B4)-8B5CF6?style=for-the-badge" alt="한국어 README 보기"></a>
- </p>
- > **Precision-Engineered Statistical Mediocrity.**
- > A Swiss-modernist design system for manufacturing calibrated, reproducible AI slop on demand.
- Every frontend agent skill on the market promises to *cure* AI slop—banishing purple-to-blue gradients, frosted glass cards, floating rocket emojis, and breathless claims about revolutionizing your workflow.
- `slop-skill` asks a more disciplined question: **what if you want the slop, engineered to specification?**
- Instead of letting a language model drift into accidental mediocrity, `slop-skill` formalizes the canonical tropes of 2022–2025 LLM SaaS landing pages into a deterministic, schema-validated, five-level design system. Turn the dial to **Level 1** for polite startup restraint, or crank it to **Level 5** for full dark-void purple neon, drifting radial blobs, `10K+ Active Users`, and five-star testimonials from fictional VPs of Engineering.
- ---
- ## 1. Project Overview & Thesis
- AI slop is not broken HTML. Broken margins and mismatched system fonts read as hurried human work. True AI slop is **hyper-polished statistical mediocrity**: the unmistakable visual convergence that appears when a model averages ten thousand Tailwind UI kits into a single centered vertical stack.
- Any single trope—an `Inter` + `Poppins` pairing, a `backdrop-filter: blur(16px)` card, a `Get Started Free` button—can survive in isolation. Recognition happens at **co-occurrence**. `slop-skill` models that co-occurrence as an explicit parameter (`level: 1..5`), backed by a seeded PRNG dealer, a slot-isolated HTML assembler, and an automated **Slop Meter** linter that measures your output on a 0–100 scale.
- ---
- ## 2. Key Features
- - **Seeded Deterministic Design Dealing:** Uses a Mulberry32 PRNG so identical `--level`, `--seed`, `--page`, and `--brand` inputs produce byte-identical JSON design specs.
- - **5-Level Slop Intensity Dial:** Cleanly scales palettes, typography, component count, animation stacks, and buzzword saturation from `Level 1` (*Subtle Slop*) to `Level 5` (*Peak Transcendent Slop*).
- - **Zero-Dependency Node.js 22+ ESM CLI:** Runs directly on native Node.js (`>= 22.0.0`) with zero `node_modules`, zero build steps, and composable Unix-style stdio pipelines.
- - **30+ Rule Slop Meter:** Evaluates any HTML document across five categories (`color`, `typography`, `layout`, `effect`, `copy`) with weighted scoring, grade classification, `--json` output, and `--min` threshold exit codes.
- - **Agent Skills & Claude Plugin Packaging:** Installable via `npx skills`, the Claude Code plugin marketplace, or a safe local symlink script (`install.sh`).
- ---
- ## 3. Slop Levels
- The five discrete intensity levels defined in [docs/CONTRACT.md](docs/CONTRACT.md):
- | Level | Label | Components | Animations | Visual & Structural Profile | Expected Slop Score |
- |---:|---|---:|---|---|---|
- | 1 | Subtle Slop | 2–3 | 0 (`[]`) | Light/soft neutral palette, subtle gradient, geometric font, clean centered hero & feature cards, minimal clutter | 0–50 (`Artisanal` / `Mild`) |
- | 2 | Mild SaaS | 3–4 | 0–1 (optional) | Soft indigo/violet accent hints, dual CTAs, modest elevation shadows, gentle buzzwords | 25–60 (`Mild` / `Full Slop`) |
- | 3 | Classic AI | 4–5 | 1 (`["float"]`) | Purple-to-blue gradient, gradient-text headline, emoji 3-column cards, sparkle pill badge, floating motion, SaaS cliché copy | 51–80 (`Full Slop`) |
- | 4 | Heavy Neon | 6–7 | 2–3 (`float`, `pulse`, `hover-scale`) | Dark void backdrop, purple/blue neon glow, glassmorphism cards, fake stats, 5-star testimonials, heavy shadows | 70–92 (`Full Slop` / `Transcendent`) |
- | 5 | Peak Transcendent Slop | 7–10 | 4–6 (all active) | Full dark neon void, multi-stop gradients, gradient text, blurred blobs, floating decorations, glassmorphism, emoji cards, heavy shadows, fake stats, 5-star testimonials, newsletter box, neon CTA, peak buzzword saturation | 81–100 (`Transcendent`) |
- ---
- ## 4. Installation
- ### Agent Skills CLI (`npx skills`)
- ```bash
- # Add all skills from the repository
- npx skills add seohuda/slop-skill
- # Target slop-skill explicitly or install globally
- npx skills add seohuda/slop-skill --skill slop-skill
- npx skills add seohuda/slop-skill --global
- ```
- ### Claude Code Plugin Marketplace
- Inside Claude Code:
- ```text
- /plugin marketplace add seohuda/slop-skill
- /plugin install slop-skill@slop-skill
- ```
- ### Local Symlink Install (`install.sh`)
- ```bash
- git clone https://github.com/seohuda/slop-skill.git
- cd slop-skill
- ./install.sh
- ```
- By default, `install.sh` symlinks `skills/slop-skill` into `${CLAUDE_SKILLS_DIR:-$HOME/.claude/skills}/slop-skill`. It refuses to overwrite existing files or conflicting directories unless `--force` is passed.
- Useful flags:
- ```bash
- ./install.sh --dry-run # Preview actions without modifying disk
- ./install.sh --target /custom/skills # Install into a custom directory
- ./install.sh --force # Replace an existing path at the target
- ```
- ---
- ## 5. Usage Guide
- ### As an AI Agent Skill
- Once installed, prompt your coding agent in Korean or English and specify your desired Slop Level (or let it default to Level 3):
- - **Korean triggers:**
- - `"AI slop 디자인을 만들어줘"`
- - `"일부러 AI스럽게 만들어줘"`
- - `"전형적인 AI 랜딩페이지 만들어줘"`
- - `"슬롭 강도 5로 디자인해줘"`
- - `"일부러 ChatGPT가 만든 것처럼 만들어줘"`
- - **English triggers:**
- - `"Make this look like AI slop"`
- - `"Build a Level 4 AI slop landing page with purple neon and glassmorphism"`
- - `"Audit index.html with the slop meter and raise it to at least 81"`
- ### Direct CLI Pipeline
- Generate a Level 5 design deal, assemble it into a self-contained HTML page, and measure the resulting Slop Score in one pipeline:
- ```bash
- node skills/slop-skill/scripts/deal.mjs --level 5 --seed 42 --brand "NovaFlow AI" \
- | node skills/slop-skill/scripts/assemble.mjs --deal - --output out/landing.html
- node skills/slop-skill/scripts/slop-meter.mjs out/landing.html --min 81
- ```
- ---
- ## 6. CLI Commands
- All scripts require **Node.js >= 22.0.0** and can also be invoked via `npm run`.
- ### `deal.mjs` — Deterministic Design Spec Generator
- ```bash
- node skills/slop-skill/scripts/deal.mjs --level <1-5> [--seed <integer>] [--page <page>] [--brand <name>] [--library <dir>]
- ```
- - `--level <1-5>`: Slop intensity level (default `3`; exits `1` if invalid or out of range).
- - `--seed <integer>`: Deterministic integer seed for the Mulberry32 PRNG (default `42`).
- - `--page <string>`: Target page archetype (`landing`, `saas`, `portfolio`, `pricing`; default `landing`).
- - `--brand <string>`: Brand name injected into `copy.brand` (default `"NovaFlow AI"`).
- - `--library <dir>`: Optional path to custom library JSON directory.
- ### `assemble.mjs` — Slot-Based HTML/CSS Assembler
- ```bash
- node skills/slop-skill/scripts/assemble.mjs --deal <path|-> [--output <path>] [--template <path>] [--css <path>]
- ```
- - `--deal <path|->`: Path to deal JSON file, or `-` (or omitted) to read JSON from `stdin`.
- - `--output <path>`: Destination HTML file path (creates parent directories automatically; defaults to `stdout`).
- - `--template <path>`: Optional custom starter HTML template overriding `skills/slop-skill/assets/starter.html`.
- - `--css <path>`: Optional custom base stylesheet overriding `skills/slop-skill/assets/system.css`.
- ### `slop-meter.mjs` — Slop Score Auditor
- ```bash
- node skills/slop-skill/scripts/slop-meter.mjs <html-file> [--min <0-100>] [--json] [--rules <dir>]
- ```
- - `<html-file>`: Target HTML file to inspect.
- - `--min <0-100>`: Minimum required Slop Score; exits with code `1` if `score < N`, otherwise `0`.
- - `--json`: Emits structured JSON (`{ file, score, grade, matchedWeight, totalWeight, passedCount, totalRules, categories, results }`).
- - `--rules <dir>`: Optional directory containing custom rule JSON files.
- ### `npm` Scripts
- ```bash
- npm test # Runs node --test across the test suite
- npm run deal -- --level 4 --seed 42 # Alias for deal.mjs
- npm run assemble -- --deal deal.json # Alias for assemble.mjs
- npm run meter -- out/landing.html # Alias for slop-meter.mjs
- ```
- ---
- ## 7. Architecture & Tooling Overview
- `slop-skill` separates **tropes** (data), **selection** (PRNG dealing), **rendering** (slot assembly), and **verification** (weighted scoring) so lower levels never accidentally inherit Level 5 neon effects:
- ```text
- skills/slop-skill/library/*.json (palettes, fonts, copy, components)
- │
- ▼
- deal.mjs + lib/prng.mjs (Mulberry32 seeded weighted sampling)
- │ JSON deal spec
- ▼
- assemble.mjs + assets/ (starter.html 9 slots + system.css + :root tokens)
- │ Self-contained HTML
- ▼
- slop-meter.mjs + rules/*.json (30+ regex/count/computed rules -> Score 0-100)
- ```
- - **9 Canonical Template Slots:** `head`, `tokens`, `hero`, `features`, `stats`, `testimonials`, `cta`, `footer`, and `extra`. Unused `<!-- SLOT:... -->` markers are stripped cleanly, and animations are emitted only when explicitly present in `deal.animations`.
- - **6 Allowed Animation IDs:** `float`, `pulse`, `blob-drift`, `gradient-shift`, `hover-scale`, `neon-glow-pulse`.
- - **Deep Dive Documents:**
- - [Anatomy of AI Slop Design](docs/slop-anatomy.md) — Taxonomy of color, typography, layout, effect, and copywriting tropes.
- - [Technical Contract](docs/CONTRACT.md) — Single Source of Truth for schemas, CLI flags, slot contracts, and scoring rules.
- ---
- ## 8. Examples
- All three showcase pages below share the exact same fictional brand brief (**NovaPulse AI (Fictional Demo)**) and deterministic seed (`--seed 127`), generated directly with `deal.mjs` and `assemble.mjs` and audited with `slop-meter.mjs`. Screenshots were captured in headless Google Chrome (`1280×1500`).
- | Level | Label | Palette | Font Pairing | Components | Animations | Measured Slop Score | Grade | HTML |
- |---:|---|---|---|---:|---|---:|---|---|
- | 1 | Subtle Slop | Slate Minimal | DM Sans / Inter | 3 | None | **19/100** (8/34 rules) | `Artisanal` | [examples/level-1/index.html](examples/level-1/index.html) |
- | 3 | Classic AI | Lavender Mist | Sora / Plus Jakarta Sans | 5 | `float` | **76/100** (26/34 rules) | `Full Slop` | [examples/level-3/index.html](examples/level-3/index.html) |
- | 5 | Peak Transcendent Slop | Aurora Violet | Syne / Plus Jakarta Sans | 9 | All 6 | **97/100** (33/34 rules) | `Transcendent` | [examples/level-5/index.html](examples/level-5/index.html) |
- ### Visual Comparison (Level 1 → Level 3 → Level 5)
- | Level 1 — Subtle Slop (`19/100`, Artisanal) | Level 3 — Classic AI (`76/100`, Full Slop) | Level 5 — Peak Transcendent Slop (`97/100`, Transcendent) |
- |---|---|---|
- | [](examples/level-1/index.html) | [](examples/level-3/index.html) | [](examples/level-5/index.html) |
- Reproduce any showcase page locally:
- ```bash
- node skills/slop-skill/scripts/deal.mjs --level 5 --seed 127 --page landing --brand "NovaPulse AI (Fictional Demo)" \
- | node skills/slop-skill/scripts/assemble.mjs --deal - --output examples/level-5/index.html
- node skills/slop-skill/scripts/slop-meter.mjs examples/level-5/index.html --min 81
- ```
- ---
- ## 9. Slop Score & Grading
- `slop-meter.mjs` evaluates an HTML document against 30+ weighted rules across five categories (`color`, `typography`, `layout`, `effect`, `copy`) and computes the normalized percentage:
- ```text
- Slop Score = Math.round((matchedWeightSum / totalWeightSum) * 100)
- ```
- | Score Range | Grade | Interpretation |
- |---:|---|---|
- | 0–25 | `Artisanal` | Suspiciously human. Almost no statistical SaaS convergence detected. |
- | 26–50 | `Mild` | Polite template discipline with restrained geometric defaults. |
- | 51–80 | `Full Slop` | Unmistakable AI-generated SaaS landing page energy. |
- | 81–100 | `Transcendent` | Peak statistical mode-collapse. Every neon glow and buzzword firing in unison. |
- ---
- ## 10. FAQ
- ### Why would anyone intentionally *add* AI slop?
- Satire, benchmarking, design-system education, and testing slop-detection heuristics. If you want to calibrate a linter or demonstrate why `linear-gradient(135deg, #8B5CF6, #3B82F6)` plus `10K+ Active Users` triggers instant fatigue, you need a reproducible baseline.
- ### Is the output deterministic?
- Yes. Passing the same `--level`, `--seed`, `--page`, `--brand`, and `--library` to `deal.mjs` produces byte-identical JSON via Mulberry32, and `assemble.mjs` renders deterministic HTML from that spec.
- ### Why zero runtime dependencies?
- An engineering tool mocking over-engineered boilerplate should not pull 400 transitive npm packages to format a CSS template string. Everything runs on standard Node.js `>= 22.0.0` ESM modules.
- ### Can I use `slop-meter.mjs` in reverse to *keep slop out* of production?
- Yes. Run `slop-meter.mjs dist/index.html --json` in CI and assert that your production marketing page stays in the `Artisanal` (`0–25`) band.
- ### How do I uninstall the local symlink?
- Remove the symlink created by `install.sh`:
- ```bash
- rm "${CLAUDE_SKILLS_DIR:-$HOME/.claude/skills}/slop-skill"
- ```
- ---
- ## 11. License
- [MIT](LICENSE)
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.