SuperAPI AI Directory
Partner中文AI newsFavorites
Skillsdesign-slop-check

design-slop-check

Design & creation

View repository

Updated 10-15 21:02

About this skill

design-slop-check — Claude Design copy AI slop check

Role reduction (P1 — docs 03 §3.3): This skill is a downstream QA gate. The **canonical owner of the AI slop pattern dictionary (English/Korean Tier 1/2 + structural anti-patterns) is now design-copywriting**, and it is proactively complied with at the copy generation stage. This skill performs post-hoc checking of result copy by referencing that dictionary, and the copy scoring criteria also use this dictionary as the SSOT. The pattern dictionary below is maintained for convenience, but the SSOT for updates is design-copywriting the "Anti-Slop Pattern Dictionary" section in the body.

Overview

Claude Design also generates copy. Even if the voice is well registered in the design system, in both English and Korean, clichéd expressions that look AI-generatedoften slip in. This skill identifies such expressions through a pattern dictionary + context inference and suggests natural alternatives.

Trigger keywords

Claude Design copy, AI slop, AI feel, clichéd expressions, just before humanize, design copy check, slop check

Check targets

Receives the following text as input.

Input type Example
Single copy text "Reimagine your workflow with AI"
Copy collection Hero, Feature, CTA, and Footer bundled at once
Entire Claude Design output Batch of text extracted from PDF/HTML
Claude Design chat capture The user selected and pasted only the copy

AI slop pattern dictionary

English clichés (Tier 1 — almost always slop)

  • "Reimagine your [X]"
  • "Unleash your [X]" / "Unleash the power of [X]"
  • "Empower your team to [X]"
  • "Transform the way you [X]"
  • "Supercharge your [X]"
  • "Revolutionize [X]"
  • "Next-generation [X]"
  • "Cutting-edge [X]"
  • "Powered by AI"
  • "Built for the future of [X]"
  • "Where [X] meets [Y]" (especially marketing headlines)

English suspect expressions (Tier 2 — depending on context)

  • "Game-changing", "Best-in-class"
  • "Seamlessly", "Effortlessly", "Intuitively"
  • "Leverage", "Synergy"
  • "World-class", "State-of-the-art"
  • "[X], simplified" / "[X], reimagined" / "[X], unleashed"

Korean clichés (Tier 1)

  • "혁신적인 [X]"
  • "차세대 [X]"
  • "재정의하는 [X]"
  • "새로운 패러다임"
  • "AI 기반의 [X]"
  • "한 차원 높은 [X]"
  • "지금까지 없던 [X]"
  • "당신의 [X]를 변화시킬"
  • "이제는 [X]의 시대"

Korean suspect expressions (Tier 2)

  • "원활하게", "손쉽게", "직관적으로"
  • "최고의", "최첨단의"
  • "강력한", "막강한"
  • "한 번에", "단숨에"
  • "더 이상 [X]에 시달리지 마세요"

Copy structure anti-patterns

  • A B C D: Word listing like "Faster. Smarter. Better. Simpler."
  • Emphasis after a colon: "[Headline]: The X that finally Y"
  • Statistical claims without numeric emphasis: "Trusted by thousands of customers"
  • Unnecessary negative → positive: "No more [X]. Just [Y]."

Korean structural slots — 3 S1 patterns (mandatory detection for headlines and copy)

The 3 most frequent in Korean slide headlines and marketing copy structural AI slop. Detected at the sentence-structure level, not via a word dictionary. A single standalone occurrence confirms an AI author, so it is classified as Tier 1 (mandatory fix).

# Pattern Detection signal [Bad example] Fix direction
1 Dash-contrast headline Dash (—)-split "X — Y" dash-contrast headline [Bad example] "From copy-paste to delegation — if you only give the goal" (dash contrast) Remove the dash and merge into one sentence or split into two sentences
2 Particle/noun-ending fragment sentence Particle/noun-ending fragment sentence ending in a particle (~은/에/로) or a noun (nominal form) [Bad example] "The key to success — automation" (particle/noun ending) Rewrite as a complete sentence including a predicate
3 "From A to B" transition formula Clichéd sentence opening the introduction with the "from X to Y" transition formula [Bad example] "From Excel to Notion, what changed" (transition formula) Start with a concrete example or evidence instead of the transition formula

Detection source: plugins/moai-writer/skills/korean-humanize/references/ai-tell-taxonomy.md D/J category. The same 3 patterns are also registered in ai-slop-reviewer and korean-humanize, so if design-slop-check misses them, the follow-up gates compensate.

Workflow

Step 1 — Copy collection

The user inputs the copy to be reviewed.

입력 옵션:
- 단일 텍스트 붙여넣기
- 파일 경로 (HTML·PDF·Markdown·텍스트)
- Claude Design URL (Export 후 공유 URL이 있다면)

Step 2 — Language detection

Automatically determines among English, Korean, and mixed. Applies each language's dictionary.

Step 3 — Pattern matching

Processing step Result
Tier 1 regex matching Almost certainly slop — auto flag
Tier 2 regex matching Needs context check — follow-up inference
Copy structure analysis Detects A B C D listing and colon-emphasis patterns
Statistical claim detection "Thousands", "tens of thousands", etc. without numerical evidence
Self-praise detection Self-asserting expressions such as "innovative", "the best"

Step 4 — Context inference

Tier 2 expressions contextdetermine whether they are slop. The following are considered.

  • Industry context: "seamlessly" is ordinary in fintech, but can be slop in SaaS marketing
  • Brand tone: "powerful" is awkward for a serious brand, but natural for a lively brand
  • User persona: "easily" is inappropriate in copy aimed at executives

When possible, ask the user about industry and brand tone (AskUserQuestion).

Step 5 — Fix suggestions

For each detected item:

원본: "Reimagine your workflow with AI-powered automation"
이유: "Reimagine your"는 거의 모든 AI 도구가 쓰는 진부 표현
대안 3개:
  1. "[제품명]은 매일 반복하는 [구체적 작업]을 자동화합니다"
  2. "[구체적 작업]에 쓰는 시간을 [숫자] 줄이는 자동화 도구"
  3. "[페르소나]를 위한 [핵심 가치] 자동화"
선호: 1번 — 동사로 시작 + 구체적 작업 명시

Step 6 — Generate Review Report

# Claude Design 카피 슬롭 검수

## 한눈에 보기

| 항목 | 결과 |
|---|---|
| 검수한 카피 수 | N |
| Tier 1 슬롭 발견 | N건 |
| Tier 2 의심 표현 | N건 |
| 구조 안티패턴 | N건 |
| 총 권장 수정 | N건 |

## Tier 1 슬롭 (필수 수정)

### Hero 헤드라인
- 원본: [원본 카피]
- 문제: [패턴]
- 대안: [3개]
- 추천: [1개]

### CTA
- 원본: ...

## Tier 2 의심 표현 (문맥 확인)

- 원본: ...
- 문제: ...
- 결정 요청: [컨텍스트 묻기]

## 구조 안티패턴

- 패턴: A B C D 나열
- 위치: Feature 섹션 3번째 카드
- 원본: "Faster. Smarter. Better. Simpler."
- 문제: 의미 없는 형용사 4개 나열
- 대안: 1개 구체적 결과로 — "고객 응답 시간을 12시간에서 30분으로 단축"

## 후속 처리 추천

1. 위 수정을 Claude Design 채팅에 다시 요청
2. 한국어 카피 자연화 → moai-writer:korean-humanize → 최종 검수
3. 영문 카피 후속 검수 → 영문 카피 베스트 프랙티스 가이드

Output Format

The review results are in the Markdown report above. The user can paste it directly into the Claude Design chat to request revisions.

Usage Examples

Example 1 — Hero + CTA

입력:
- Hero: "Reimagine the way your team collaborates"
- CTA: "Get started for free"

결과:
- Hero: Tier 1 슬롭 — "Reimagine the way" 패턴
  대안:
  1. "팀이 매일 쓰는 N개 도구를 하나로 합칩니다"
  2. "회의를 [숫자] 줄이는 협업 도구"
  3. "[페르소나]를 위한 협업 워크플로우"
- CTA: 평범 — "Get started for free"는 슬롭 아님, 유지

권장 수정:
- Hero를 1번으로 교체 → 다시 Claude Design에서 시안 갱신

Example 2 — Korean Marketing Page

입력:
- Hero: "혁신적인 AI 기반의 차세대 마케팅 자동화"
- 서브: "당신의 마케팅을 한 차원 높여 줄 솔루션"
- Feature: ["빠르게", "쉽게", "정확하게", "강력하게"]

결과:
- Hero: Tier 1 슬롭 3개 ("혁신적인", "AI 기반의", "차세대") 동시 사용
  대안:
  1. "주간 보고서 자동 생성·발송 — 30분 작업이 0분"
  2. "[고객 사례] 처럼 채널 광고 운영 시간 [N시간 → M시간]"
  3. "마케팅 자동화로 [구체적 결과]"
- 서브: Tier 1 슬롭 ("한 차원 높여")
- Feature: 구조 안티패턴 — A B C D 부사 나열
  대안: 각각 구체적 결과·수치로 교체

Example 3 — Mixed (English + Korean)

입력:
- 영문 Hero + 한국어 서브
- 영문 CTA + 한국어 푸터

결과: 언어별 사전 적용, 각각 분석

Precautions

Do

  • Tier 1 is almost always a revision recommendation — usable with self-censorship
  • Tier 2 requires context checking — do not make arbitrary judgments
  • Alternative copy is specific results and figuresto include — clichéd expressions are abstract and therefore slop
  • As a follow-up, chaining moai-writer:korean-humanize is recommended (for Korean results)

Don't

  • Do not advise against all English copy on the grounds of "AI slop" — some expressions are industry standard
  • Do not mistake the user's brand's intended tone for slop — check context first
  • Do not automatically batch-edit — always apply after user review

Related Skills

Skill When to Use
moai-writer:korean-humanize Follow-up: Korean copy naturalization (SSOT of 10 major categories × 40+ AI tell patterns)
moai-coworker:ai-slop-reviewer Follow-up: general AI slop review of all text outputs
moai-designer:design-brief Prerequisite: refine the brief so better copy is produced
moai-marketer:marketing-campaign-planner Auxiliary: review tone and message consistency of campaign copy
moai-marketer:content-copywriting Alternative: generate copy directly outside Claude Design