SuperAPI AI Directory

cs-channel-message

General & other

View repository

Updated 10-15 21:38

About this skill

Automatic message generation per channel (Commerce Channel Message)

Overview

This is a skill that automatically generates a total of 15 messages—3 channels (search, advertising, CRM) × 5 types each—with different expressions per channel according to the NCM framework (Need → Channel → Moment → Message → CTA) based on JTBD and personas.

Data sources:

  • Seasonal/season-specific keyword calendar (Moment determination) — public trend data such as Naver DataLab or seasonal keyword materials attached by the user
  • Ad keyword performance data (CTA optimization) — performance report files exported from advertising platforms (Naver Search Ads, Coupang Ads, etc.)

Data integration: If there is a connected commerce MCP connector, query it directly; if not, generate the same output using export files/screenshots attached by the user or Claude's own analysis.

Trigger keywords

Channel message, NCM framework, search ad copy, banner ad copy, CRM message, KakaoTalk notification, email subject, SNS ad copy, per-channel copy, ad copy branching, seasonal message, customer touchpoint message, 15 types of copy, ad headline, Meta ads, TalkTalk response, push notification, email sequence, cart abandonment, repurchase message, app push planning, retention push, discount push, push 4 principles, Timely Personal Actionable, gamification push, numeric push vs branding push

3 operation modes

This skill is a single-channel message engine in which the mode is automatically selected from natural language.

Mode Example trigger utterance Output Detailed reference
Channel-branch message mode (core) "15 types of messages per channel", "Apply NCM" Search/advertising/CRM 5 types each = 15 branch messages NCM framework body below
Operational copy mode "Meta ad headline", "TalkTalk auto-response", "3 cart abandonment emails", "repurchase notification" Copy tailored to channel, length, and tone + 2-3 A/B variations references/ad-copy.md · talk-copy.md · push-copy.md · content-email-sequence.md
App push planning mode "Create app push copy", "3 retention push options", "gamification push" 4-principle matching + 3-element check + 3 variations + click-through rate prediction references/push-planning.md

The mode is determined by the utterance, not by a CLI flag. "Create app push copy" enters push planning mode, "search ad copy" enters operational copy mode, and "15 types of messages per channel" enters core mode.

NCM framework

Step Description Input source
N — Need Define the customer's core needs JTBD 9개 (commerce-jtbd-persona)
C — Channel Select message delivery channel Enter priority channel
M — Moment Optimal contact timing Seasonal keyword calendar (DataLab·attachments)
M — Message Write channel-specific tailored messages JTBD + Persona
CTA Channel-specific call-to-action phrases Ad keyword performance report

Key constraints (HARD): Even for the same need, must branch into different expressions per channel.


Workflow

Input slots

Item Required Example
JTBD result Required commerce-jtbd-persona --mode jtbd output
Persona Recommended commerce-jtbd-persona --mode persona output
Priority channel Required Search ads, banner ads, CRM (select at least 1)
Season/Timing Optional "Spring season", "Year-end promotion" (default: current month)

Message generation order

1. 시즌 키워드 정리 (카테고리, 현재 월 기준 — 데이터랩 조회 결과 또는 첨부된 시즌 키워드 자료)
   → 현재 시즌 핵심 키워드 + 접촉 최적 시점

2. 광고 키워드 성과 검토 (광고 플랫폼 내보내기 리포트 첨부 시)
   → 채널별 광고 키워드 성과 → CTA 문구 최적화 힌트

3. NCM 매핑:
   [N] JTBD 우선순위 1위 니즈 선정
   [C] 채널 3개 분류: 검색(네이버/쿠팡) / 광고(배너/SNS) / CRM(카카오/이메일/SMS)
   [M] 시즌 키워드 캘린더 기반 최적 시점 설정
   [M] 채널별 다른 표현으로 메시지 5종씩 작성
   [CTA] 키워드 성과 데이터 기반 채널별 CTA 최적화

ai-slop-reviewer auto-chaining (HARD)

Immediately after generating 15 message types moai-coworker:ai-slop-revieweris automatically chained.

Review items:

  • Remove AI-pattern messages ("innovative", "amazing", "best" clichés)
  • Match natural tone per channel (search: keyword-style / CRM: conversational / SNS: impactful)
  • Adjust to actual e-commerce ad tone

Usage example

"/cs-channel-message 비건 스킨케어, JTBD+페르소나 첨부, 검색+CRM 우선"
→ 검색광고 5종: 키워드 기반 헤드라인
   배너광고 5종: 비주얼 후킹 메시지
   CRM 5종: 카카오/이메일/SMS 맞춤 문구
   → ai-slop-reviewer 자동 검수

"/cs-channel-message 반려견 관절 간식, 봄 시즌, 쿠팡 광고 우선"
→ 시즌 키워드 캘린더 반영 + 15종 분기 메시지

Output format

{
  "product": "{상품명}",
  "season_moment": "{현재 시즌 + 최적 접촉 시점}",
  "ncm_framework": {
    "need": "{핵심 니즈 1줄}",
    "channels": ["검색", "광고", "CRM"],
    "moment": "{최적 접촉 시점}"
  },
  "messages": {
    "search": [
      {
        "id": "S1", "channel": "네이버 검색광고",
        "headline": "메시지 제목 (15자 이내)",
        "description": "설명 문구 (45자 이내)",
        "cta": "CTA 문구",
        "keyword_trigger": "타겟 키워드"
      },
      {"id": "S2", "channel": "쿠팡 검색광고", "...": "..."},
      {"id": "S3", "channel": "네이버쇼핑 검색", "...": "..."},
      {"id": "S4", "channel": "카카오 검색", "...": "..."},
      {"id": "S5", "channel": "구글 검색광고", "...": "..."}
    ],
    "ad": [
      {"id": "A1", "channel": "인스타그램 배너", "headline": "...", "body": "...", "cta": "..."},
      {"id": "A2", "channel": "페이스북 배너", "...": "..."},
      {"id": "A3", "channel": "카카오 배너", "...": "..."},
      {"id": "A4", "channel": "유튜브 인피드", "...": "..."},
      {"id": "A5", "channel": "스마트스토어 배너", "...": "..."}
    ],
    "crm": [
      {"id": "C1", "channel": "카카오 알림톡", "message": "...", "cta": "..."},
      {"id": "C2", "channel": "이메일 제목", "subject": "...", "preview": "..."},
      {"id": "C3", "channel": "SMS", "message": "...(90자 이내)"},
      {"id": "C4", "channel": "카카오 친구톡", "...": "..."},
      {"id": "C5", "channel": "앱 푸시", "title": "...", "body": "..."}
    ]
  },
  "channel_differentiation_check": "PASS — 3채널 표현 상이 확인",
  "slop_review": {
    "status": "passed",
    "changes_made": 2,
    "notes": "ai-slop-reviewer 검수: '놀라운 효과' → '피부과 내원 전에' 수정"
  }
}

Quality checklist

  • 15 types completed: 5 each for search, ads, and CRM, 15 total
  • Different expressions per channel: The same need branches in different ways per channel
  • ai-slop-reviewer review traces: Includes slop_review block
  • Includes CTA: Each message has a call-to-action phrase suited to the channel's characteristics

Related skills

Chaining order: commerce-jtbd-persona --mode persona → cs-channel-message → commerce-integrated-strategy

  • commerce-jtbd-persona — JTBD+Persona (previous step, Need definition input)
  • commerce-product-naming — Confirmed product name (use product name in messages)
  • commerce-integrated-strategy — Comprehensive 1-page strategy including channel messages (next step)
  • commerce-message-compliance-kr — Pre-send legal gate (push·notification talk·email ads)
  • moai-marketer:content-sns-content — SNS content standalone deep-dive work
  • moai-coworker:ai-slop-reviewer — AI pattern review (auto-chaining)
  • moai-writer:korean-humanize — Remove Korean AI markers (after slop review, required)

When not to use this skill

  • Domain-unspecific single-purpose copy: design-copywriting use
  • General advertising copy outside e-commerce: moai-marketer:content-copywriting use
  • Detail page 13-section copy: commerce-detail-page-copy use
  • Email newsletter body: moai-marketer:content-newsletter or moai-marketer:content-email-sequence use
  • Full ad campaign planning: moai-marketer:marketing-campaign-planner use
  • Actual ad execution·campaign operations: Directly managed on the ad platform

Operational copy mode (ads, TalkTalk, push, email)

Enter here when you need operational copy for a single channel, such as "Create a Meta ad headline," "TalkTalk auto-reply template," or "3 cart abandonment recovery emails." Beyond store detail pages, it produces the copy constantly needed for ads, CS responses, notifications, and emails while adhering to constraints on length and tone by channel and format.

Copy categories

  1. Ad copy — Search (Naver, Google), display/social (Meta, Kakao Moment), video (YouTube, Reels) → references/ad-copy.md
  2. TalkTalk/CS responses — Auto-replies, FAQ, 1-star review apology, refund/exchange handling → references/talk-copy.md
  3. Push notifications — New products, promotion start/end, cart recovery, repurchase → references/push-copy.md
  4. Email/AlimTalk — Order confirmation, shipping, review requests, repurchase coupons, newsletters, cart abandonment recovery → references/email-sequence.md

Workflow

  1. Identify copy category + channel (length and tone differ by channel)
  2. Collect context (product, target, purpose, tone, length limits)
  3. Produce copy by channel (references above)
  4. 2-3 A/B variants (A: direct/functional, B: emotional/lifestyle, C: urgent/story)
  5. moai-coworker:ai-slop-reviewer Auto-chaining

Length constraints by channel (cheat sheet)

Channel Headline Body Notes
Naver Search Ads 15자 45자 Keyword matching matters
Google Search Ads 30자 × 3 90자 × 2 RSA format
Meta Ads 40 characters (mobile) 125 characters (risk of truncation) Visual-first
Kakao Moment 20자 60자 Mobile only
Kakao AlimTalk 16자 (header) 1,000자 (body) Separate consent for ads
TalkTalk auto-reply No limit 1,000자 Polite tone
Push notifications 30 characters (title) 120 characters (body) Short and strong
Email subject 50자 - 25 characters is the mobile truncation limit

Operational copy mode precautions

  • Kakao AlimTalk: Advertising messages require separate consent. Transactional notifications are not ads.
  • Email advertising: Information and Communications Network Act — sender information and unsubscribe link required.
  • Absolute expressions such as "lowest price" may violate the Act on Fair Labeling and Advertising.
  • Push can be blocked by OS settings → limit frequency (2-3 times per week or less).

App push planning mode (4 principles, 3 elements, 3 variants)

"Create app push copy," "3 retention push copy variants," "Gamification push" — enter here when planning app pushes from the ground up rather than sending them blindly. It covers in depth the lock screen, summary display, and gamification tone and manner unique to app pushes. Detailed frameworks, matrices, 30+ brand references, and click-through rate prediction formulas are references/push-planning.md See.

Inspection order

[Step 1] 4원칙 매칭 — WHY(목적 4분류) / WHEN(트리거) / WHO(세그먼트) / HOW(톤앤매너)
         → 하나라도 미답이면 발송 차단
[Step 2] 3요소 자가 점검 — Timely / Personal / Actionable 모두 PASS
[Step 3] 비즈니스 타입 톤앤매너 매핑 — D2C(정체성·감성) / 마켓(숫자·CTA) / 결제·구독(절제·기능)
[Step 4] 카피 변형 3안 — 오늘만 vs 매일 / 누구나 vs 너에게만 / 숫자·게이미피케이션·브랜딩
[Step 5] 클릭률 예측 (기본 1.5% + 3요소·개인화·게이미피케이션 가산, 야간 발송 감산)

Input slots

Item Required Example
Sending purpose Required Service exposure / discounts·events / awareness / retention
Business type Required D2C / marketplace / payment·subscription
User segment Required All members / not accessed in last 7 days / cart abandoned 24h / 60 days after purchase
Planned send time Required 2026-05-18 12:30 (lunch golden hour)
Core benefits·content Required Vegan serum 20% discount
Tone and manner preference Recommended Numbers / gamification / branding

Push planning follow-up: right before sending, moai-seller:commerce-message-compliance-krpasses through the Network Act gate (prohibition of advertising sends at night 21:00-08:00, opt-out, sender information).


Integrated Guide to Advertising Psychology

6 Psychological Triggers

When generating NCM messages, explicitly apply 1-2 triggers to every message:

Trigger Mechanism Message example
Trust Multiplier of the value equation — if 0, value is also 0 "Real-name reviews ★4.8 / 3,247 cases"
Loss Aversion Losses are felt 2.5 times more strongly than gains "Closes today at midnight · last 3 left"
Social Proof Cues from others' behavior — differentiation by motivation "100,000 people chose it" (mass) / "500 expert baristas" (experts)
Cognitive Ease If it's complicated, they won't buy "3-click purchase · KakaoPay"
Identity Buy "the me who uses this" "Chosen by those who think seriously about coffee"
Anchoring The first number seen becomes the standard "List price 98,000 → 49,000 won"

Trust = multiplier: Even if all other triggers are strong, if trust is 0, value = 0. Trust elements (real-name reviews·verification·refund guarantee·brand consistency) are always included in the message.

Psychological State Matrix by Channel

Each channel is where people in completely different psychological states are:

Channel Psychological state Strong motivation Message strategy Precautions
Meta (Instagram·Facebook) Passive discovery Psychological needs·identity Hook·emotional trigger·visual Never touch the 48-72 hour learning period. CAPI required
Google (Search·YouTube) Active search Functional need·hot target Classification by keyword intent (informational/navigational/purchase) 70%+ of ad spend on purchase-intent keywords. Negative keywords ('free'·'DIY')
Naver Start of information search Functional need·blog reviews Blog·review ecosystem + SEO Blogs·reviews before ad spend. Korean consumer search→blog→purchase pattern
Kakao Super target 90%+ view rate Identity·repurchase KakaoTalk messages·AlimTalk·identity ROI overwhelmingly higher for retaining existing customers than acquiring new ones
Coupang Search area + non-search area simultaneously Immediate purchase Separate search (conversion message) vs non-search (impression message) 167x CPM difference → different messages per area. Linked with moai-seller:commerce-marketplace-coupang-ads

Guide to applying 9 cognitive biases

Pair skill moai-marketer:marketing-campaign-plannersame 9 types as (based on the standard Kahneman behavioral economics catalog).

Bias Message application pattern
Framing "10% failure rate" ❌ → "90% success rate" ✅ (gain frame). Loss frame is stronger for loss-aversion motivation
Halo effect Premium package design = product looks higher quality. Endorsement by a trusted influencer
Compromise effect Among 3 price options, middleplace the one you want to sell (low/mid/high price)
Present bias "Order today → arrives tomorrow", "right now" — emphasize immediate results
Sunk cost Free trial + data accumulation → "if I leave, all this disappears" → paid conversion
Zero risk "Refund within 30 days, no questions asked" (zero = stronger than 90% refund)
Choice overload Keep options to 3 or fewer + emphasize one as 'recommended' (Iyengar 24→6 jams, 10x difference)
Default effect Align with the default selection — "recommended option pre-selected" / "auto-renewal ON by default" / pre-checked consent to long-term retention terms
Mere exposure Liking what's familiar — the psychological basis of retargeting campaigns

Bias combination effect: scarcity (only 3 in stock) + loss aversion (if you don't buy now, it's gone) + social proof (47 people viewing) — placing them on one page produces an effect far exceeding the strength of each individually.

Add advertising psychology metadata to NCM output

{
  "channel": "메타 (인스타 피드)",
  "message": "...",
  "triggers": ["손실 회피", "사회적 증거"],
  "biases": ["프레이밍", "단순 노출 (리타기팅)"],
  "target_temperature": "웜 (광고 50% 시청자)",
  "rationale": "수동적 발견 채널 → 후크 + 감정 촉발 + 마감 긴급성"
}

Full-stack of Korean 30+ brand messages by AARRR stage (reinforced in v2.7.0)

A catalog of messages from 30+ Korean D2C and service brands, cross-referencing the NCM framework with the 5 AARRR stages (Acquisition, Activation, Retention, Revenue, Referral). Each stage has different tone, CTA, and psychological triggers.

Acquisition — New awareness and acquisition

Brand Channel Message pattern Psychological trigger
Toss KakaoTalk·App push "Transfer with Toss for a 100 million won draw" Curiosity·Gamification
Karrot Market Search ads "Trades close by in my neighborhood" Belonging·Proximity
29CM Instagram ads "A select just for you" Identity·Curation
Duolingo App download "Just 5 minutes to start a foreign language" Cognitive ease
Inflearn Naver ads "Courses by working professionals 50% off open" Social proof

Activation — First action and onboarding

Brand Channel Message pattern Psychological trigger
Baemin App push "Free delivery on your first order (24 hours only)" Loss aversion·Urgency
Coupang Friend Talk "Extra 10% immediately upon Rocket Wow sign-up" Loss aversion
Yanolja Email "5,000 won instant discount on your first booking" Instant reward
Class101 App push "Notification to start learning the course you purchased" Actionable
KakaoBank Friend Talk "Create an account → done in 1 minute" Cognitive ease

Retention — Encouraging revisits and repurchases

Brand Channel Message pattern Psychological trigger
Ohprint.me App push "14 days since your last design — continue working on it" Sunk cost
Publy Friend Talk "Continue the article you left off reading last week" Sunk cost·Self-consistency
Collins Email "60 days with your perfume — time to meet again" Identity
Inflearn App push "14 days since your last study — continue listening to the course" Sunk cost
Yanolja Friend Talk "The accommodation you saved dropped by 50,000 won" Loss aversion·Anchoring
Netflix App push "Next season revealed — continue watching" Sunk cost·identity
Rounds App push "🎯 Day 5 of attendance — next reward ★★★" Gamification
Duolingo App push "The owl is sad. Learning streak at risk!" Loss aversion·emotion
Remember Friend Talk "28 business cards last week — time to organize" Sunk cost

Revenue — immediate sales·conversion

Brand Channel Message pattern Psychological trigger
Ably App push "20% today only — midnight deadline" Loss aversion·urgency
Zigzag App push "30 minutes after work only 50%" Time-limited·golden time
Musinsa App push "Last 1 in size, right now" Scarcity·loss aversion
Always Friend Talk "Today's price -₩2,000 from yesterday" Anchoring·immediacy
Coupang Eats App push "Lunch within 5 minutes — order closes in 30 minutes" Time golden time

Referral — recommendation·viral

Brand Channel Message pattern Psychological trigger
Toss KakaoTalk·app push "Invite a friend → ₩5,000 instantly" Instant reward
Karrot Market Friend Talk "Notify neighborhood neighbors" Social stimulation
Uber Eats Email "Earn ₩7,000 with a friend invite code" Social proof·reward
KakaoPay Friend Talk "Transfer → have your friend receive it via KakaoPay too" Simple exposure·spread
Yanolja App push "Travel review → ₩5,000 instantly" Social proof·reward

3-element checklist (regardless of stage)

Right before sending an NCM message, self-check the following 3 elements (same criteria as the 3 elements of app push planning mode):

  • Timely: When the message is received, is it a situation where immediate action is possible? (Commute 7-9 AM / lunch 12-1 PM / after work 6-8 PM / personal time 10 PM-12 AM)
  • Personal: Does it reflect the recipient's segment·name·purchase history? (All < segment < 1:1)
  • Actionable: Is the next action clear? (Tap location + time + benefit)

When all three elements are met, average click-through rate 1.5% → 4-8% (2-5x increase).

Recommended sending frequency by AARRR stage: Acquisition 1-2 times per week / Activation triggered within the first 7 days / Retention based on behavior triggers (7, 30, 60, and 90 days after purchase) / Revenue per campaign / Referral triggered immediately after the action.