Co Scientist Literature Review

Literature Review

Systematic literature search, screening, and evidence synthesis.

Use This Skill When

• Searching for prior research on a topic.
• Conducting a PRISMA-compliant systematic review.
• Screening and evaluating source quality.
• Synthesizing findings across multiple studies.
• Identifying research gaps.

Workflow

1. Define search strategy:

   • Keywords and Boolean operators
   • Target databases (PubMed, Scopus, Web of Science, Google Scholar)
   • Inclusion/exclusion criteria
   • Date range and language filters

2. Execute search and screen results:

   • Title/abstract screening
   • Full-text screening
   • Quality assessment (risk of bias)

3. Extract and synthesize:

   • Data extraction table
   • Thematic analysis or narrative synthesis
   • Gap identification

4. Generate PRISMA flow diagram data

5. Save all outputs to files

Deliverables

• report.md: synthesis narrative with key findings.
• results/search-strategy.md: documented search methodology.
• results/screening-table.csv: inclusion/exclusion decisions.
• results/extraction-table.csv: extracted data from included studies.
• figures/prisma-flow.md: PRISMA flow diagram data.

Available Tools (MCP)

External tools available via ToolUniverse MCP server. Falls back to Python requests + public REST APIs when MCP is unavailable.

• Read references/prisma-guide.md when conducting PRISMA-compliant systematic reviews.

Quality Gates

• Search strategy is documented and reproducible.
• Inclusion/exclusion criteria are explicit.
• At least 3 sources are cross-validated for key claims.
• Single-source findings are marked with ⚠️.
• Research gaps are identified with supporting evidence.

If any gate fails: identify the specific failing check, fix the issue, and re-validate before proceeding.

Gotchas

• Google Scholar の検索結果は網羅性が低い。必ず PubMed/Scopus と併用すること
• 単一ソースの情報は ⚠️ マークを付け、断定的な結論に使用しないこと
• プレプリント（bioRxiv, arXiv）は査読前であることを明記すること
• 検索語の日英バイリンガル展開を行うと、非英語圏の重要な研究を見落とさない
• スクリーニング結果は逐次ファイルに保存すること。コンパクションで中間結果が消失する

Validation Loop

1. 検索戦略とスクリーニング結果を生成
2. チェック:
   • 検索語が研究目的をカバーしているか
   • 2つ以上のデータベースを使用しているか
   • 単一ソースの主張に ⚠️ が付いているか
3. 不合格なら検索戦略を修正して再実行
4. 合格後のみ合成フェーズへ進む