claude-skills-journalism MCP — 사용 사례, 설치 & 라이브 데모

왜 쓰나요

핵심 기능

Claim verification with source-linking discipline
FOIA request drafting (federal and state templates)
Data journalism: CSV cleanup, exploratory analysis, chart drafts
Academic writing: literature review, citation formatting
Interview question scaffolding
Anti-hallucination patterns (require URLs for every factual claim)

라이브 데모

실제 사용 모습

claude-skills-journalism-skill.replay ▶ 준비됨

0/0

설치

클라이언트 선택

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

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

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

~/.cursor/mcp.json · .cursor/mcp.json

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

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

VS Code → Cline → MCP Servers → Edit

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

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

~/.codeium/windsurf/mcp_config.json

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

Claude Desktop과 같은 형식. Windsurf 재시작 후 적용.

~/.continue/config.json

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

Continue는 맵이 아닌 서버 오브젝트 배열 사용.

~/.config/zed/settings.json

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

context_servers에 추가. 저장 시 Zed가 핫 리로드.

claude mcp add claude-skills-journalism-skill -- git clone https://github.com/jamditis/claude-skills-journalism ~/.claude/skills/claude-skills-journalism

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

사용 사례

실전 활용법: claude-skills-journalism

How to verify every factual claim in a draft before publication

👤 Reporters and editors fact-checking copy on deadline ⏱ ~45 min intermediate

언제 쓸까: Draft is done; publication in 2 hours; you need to confirm each factual claim has a source.

사전 조건

Skill installed — git clone https://github.com/jamditis/claude-skills-journalism ~/.claude/skills/journalism-skills

흐름

Extract claims

Read draft.md. Extract every factual claim (dates, numbers, quotes, causal statements). Number them.✓ 복사됨

→ Numbered list of distinct factual claims
Source each

For each claim, either link the source URL or flag 'NEEDS VERIFICATION'.✓ 복사됨

→ Per-claim source or NEEDS VERIFICATION flag
Re-check flagged

For each NEEDS VERIFICATION, search for a source. If none, mark for removal or softening.✓ 복사됨

→ Resolution per claim

결과: A draft where every factual claim has either a source or a documented absence.

함정

Claude invents plausible-looking source URLs — Require the skill to fetch each URL and quote the sentence supporting the claim

함께 쓰기: notebooklm-skill

Draft a targeted FOIA request

👤 Investigative reporters requesting government records ⏱ ~25 min intermediate

언제 쓸까: You know the agency and the type of record; you want a tight, scoped request that isn't rejected for vagueness.

흐름

Scope

FOIA draft: EPA, emails between Region 3 administrators and Company X between Jan-Jun 2024 mentioning 'groundwater'. Tight scope to avoid rejection.✓ 복사됨

→ Request with specific keywords, date range, agency, and fee-waiver request
Exemption pre-check

Which FOIA exemptions is this likely to trigger, and how should I phrase to minimize them?✓ 복사됨

→ Exemption-by-exemption analysis

결과: A FOIA request likely to produce records rather than rejection.

함정

Too broad triggers fees/rejection — Always set a date range and keyword list; drop buzzwords

Explore a government dataset for a data-journalism story

👤 Data reporters looking for story angles in a new dataset ⏱ ~60 min intermediate

언제 쓸까: You downloaded a CSV/Excel from an agency; you don't yet know what's interesting.

흐름

Profile

Load data/inspections.csv. Column types, missing counts, top values per categorical, basic numeric stats.✓ 복사됨

→ Data profile output
Find angles

Suggest 5 angles an investigative reporter might find here, each with the specific query/filter that surfaces it.✓ 복사됨

→ Story angles with concrete queries
Draft chart

Chart the most compelling angle. Output ggplot / Python matplotlib code.✓ 복사됨

→ Chart code + description

결과: 2-3 publishable angles, one chart-ready.

함정

Treating correlation as causation in the pitch — Ask Claude to caveat every causal-sounding claim

Draft a literature review for an academic paper

👤 Academic writers compiling related work ⏱ ~90 min advanced

언제 쓸까: You have 20-50 papers and need to synthesize them into a coherent review.

흐름

Cluster papers by theme

Read these 30 abstracts and cluster by theme. Flag methodological disagreements.✓ 복사됨

→ Thematic clusters + tension points
Draft review

Draft a 1500-word lit review organized by theme, with proper inline citations (author, year).✓ 복사됨

→ Draft with correct citations

결과: A draft lit review ready for subject-matter refinement.

함께 쓰기: notebooklm-skill

조합

다른 MCP와 조합해 10배 효율

claude-skills-journalism-skill + notebooklm-skill

NotebookLM grounds claims in uploaded primary sources

Upload these primary sources to NotebookLM, then use journalism skill to verify every claim in the draft against them.✓ 복사됨

도구

이 MCP가 노출하는 것

도구	입력	언제 호출	비용
Claim extraction	text	First fact-check pass	Claude tokens
Source linking	claims	Second pass	Claude tokens + web search
FOIA drafting	agency, scope	Before submitting a request	Claude tokens
Dataset exploration	file path	New dataset arrives	Claude tokens
Lit review	abstracts	Academic writing	Claude tokens (high)

비용 및 제한

운영 비용

API 쿼터: None
호출당 토큰: Varies; lit reviews 50k+
금액: Free
팁: Chunk large verification jobs by story section.

보안

권한, 시크릿, 파급범위

자격 증명 저장: No credentials

데이터 외부 송신: Drafts and sources sent to Claude for inference. Don't paste sensitive source material (e.g. named confidential sources) into prompts.

Confidential sources: do not include identifying details in prompts. Treat Claude as a public-adjacent collaborator for sensitive investigations.

문제 해결

자주 발생하는 오류와 해결

Claude cites URLs that 404

Require the skill to fetch each cited URL and quote the sentence supporting the claim; reject unverifiable citations.

FOIA requests rejected as overbroad

Always scope by date range and specific keywords; drop general 'any records related to'

대안

claude-skills-journalism 다른 것과 비교

대안	언제 쓰나	단점/장점
Manual fact-check	Sensitive investigations where LLM involvement is risky	Slower, but full source control

claude-skills-journalism