startup-skill MCP — 사용 사례, 설치 & 라이브 데모

왜 쓰나요

핵심 기능

Idea validation framework (demand, willingness to pay, technical feasibility)
JTBD-style interview prompts and analysis
Competitive scans with source-backed claims
Business model sketches (pricing, acquisition, unit economics)
Lightweight planning — 30/60/90, OKR drafts
Adversarial red-team prompts (why this fails)

라이브 데모

실제 사용 모습

startup-skill.replay ▶ 준비됨

0/0

설치

클라이언트 선택

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "startup-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/ferdinandobons/startup-skill",
        "~/.claude/skills/startup-skill"
      ],
      "_inferred": true
    }
  }
}

Claude Desktop → Settings → Developer → Edit Config 열기. 저장 후 앱 재시작.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "startup-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/ferdinandobons/startup-skill",
        "~/.claude/skills/startup-skill"
      ],
      "_inferred": true
    }
  }
}

Cursor는 Claude Desktop과 동일한 mcpServers 스키마 사용. 프로젝트 설정이 전역보다 우선.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "startup-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/ferdinandobons/startup-skill",
        "~/.claude/skills/startup-skill"
      ],
      "_inferred": true
    }
  }
}

Cline 사이드바의 MCP Servers 아이콘 클릭 후 "Edit Configuration" 선택.

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "startup-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/ferdinandobons/startup-skill",
        "~/.claude/skills/startup-skill"
      ],
      "_inferred": true
    }
  }
}

Claude Desktop과 같은 형식. Windsurf 재시작 후 적용.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "startup-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/ferdinandobons/startup-skill",
        "~/.claude/skills/startup-skill"
      ]
    }
  ]
}

Continue는 맵이 아닌 서버 오브젝트 배열 사용.

~/.config/zed/settings.json

{
  "context_servers": {
    "startup-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/ferdinandobons/startup-skill",
          "~/.claude/skills/startup-skill"
        ]
      }
    }
  }
}

context_servers에 추가. 저장 시 Zed가 핫 리로드.

claude mcp add startup-skill -- git clone https://github.com/ferdinandobons/startup-skill ~/.claude/skills/startup-skill

한 줄 명령. claude mcp list로 확인, claude mcp remove로 제거.

사용 사례

실전 활용법: startup-skill

How to pressure-test a startup idea before you build

👤 First-time founders and side-project builders ⏱ ~60 min beginner

언제 쓸까: You have an idea that excites you and you want to hear the case against it before committing a quarter.

사전 조건

Skill installed — git clone https://github.com/ferdinandobons/startup-skill ~/.claude/skills/startup-skill

흐름

Frame the idea

Validate this idea using startup-skill: 'AI-powered observability for Rails apps, targeting 5-20 dev teams.'✓ 복사됨

→ Structured breakdown: customer, problem, willingness-to-pay, existing alternatives
Adversarial review

Now play devil's advocate — give me 5 reasons this fails in 18 months.✓ 복사됨

→ Specific, falsifiable failure modes (not generic)
Draft 5 validation interviews

Write 5 customer interview questions to disconfirm my riskiest assumption.✓ 복사됨

→ Non-leading, behavior-probing questions

결과: A go/no-go decision backed by specific risks and a kill-criteria list.

함정

Claude defaults to cheerleading — Explicit framing: 'act as a skeptical senior VC who has seen this category fail twice'
Interview questions leak the answer — Ask for Mom Test-style questions only

함께 쓰기: goose-skill

Run a competitive scan with sourced claims

👤 Founders preparing a pitch or positioning doc ⏱ ~40 min intermediate

언제 쓸까: You need to know who's in your category and where the gaps are.

흐름

Scope the category

Use startup-skill competitive scan for 'error monitoring for Rails apps'. Require URLs for every feature claim.✓ 복사됨

→ Vendor table with linked source per feature
Identify the whitespace

What positioning claim is unowned in this table?✓ 복사됨

→ Specific angle backed by the table

결과: A positioning angle you can defend in an investor meeting.

함정

Hallucinated features or pricing — Reject any claim without a URL; re-check URLs are real by asking Claude to fetch them

함께 쓰기: goose-skill

Draft a 30/60/90-day plan for a new role or initiative

👤 Founders and early employees needing a plan, fast ⏱ ~25 min beginner

언제 쓸까: New role, new team, or new product direction.

흐름

Seed with context

30/60/90 for a founding PM joining a 10-person dev tools startup. Primary goal: ship the self-serve signup.✓ 복사됨

→ Milestones per phase with ownership and dependencies
Pressure-test

What's over-scoped? What's the one thing that if we skip ruins the plan?✓ 복사됨

→ Specific cuts + the single critical path item

결과: A realistic plan with a stated critical path and cut candidates.

조합

다른 MCP와 조합해 10배 효율

startup-skill + goose-skill

Startup validates the bet, Goose executes GTM

Validate the idea with startup-skill; if it passes, generate ICP and outbound with goose-skills.✓ 복사됨

startup-skill + ai-marketing-claude-code-skill

Startup sets positioning, marketing skill drives campaigns

Run positioning workshop with startup-skill, then feed to ai-marketing skill for campaign planning.✓ 복사됨

도구

이 MCP가 노출하는 것

도구	입력	언제 호출
Idea validation	-	Pre-commit to a direction
JTBD interviews	-	Customer research
Competitive scan	-	Positioning and pitch prep
Business model sketch	-	Before setting prices
30/60/90 & OKRs	-	Starting something new

비용 및 제한

운영 비용

API 쿼터: None
호출당 토큰: 10-30k for a full validation pass
금액: Free
팁: Batch questions into one session per topic to reuse context.

보안

권한, 시크릿, 파급범위

자격 증명 저장: No credentials

데이터 외부 송신: Competitive scan uses Claude's web search; no persistent egress

Confidential ideas shared with Claude are subject to Anthropic's data policy — consider a workspace with no-training setting.

문제 해결

자주 발생하는 오류와 해결

Validation output feels like a pep talk

Explicitly ask for the adversarial mode and require failure modes before strengths

Competitive scan invents features

Require URLs for every claim; reject unsourced ones

대안

startup-skill 다른 것과 비교

대안	언제 쓰나	단점/장점
goose-skill	You're past validation and need GTM execution	Less idea-stage depth

startup-skill