deep-research

name	deep-research
description	Multi-source, citation-backed research with Korean reports and traceable evidence. Triggers: research topic, investigate, deep research, multi-source analysis.

deep-research

Usage

Use this skill for complex investigations that need multiple sources, cross-checking, synthesis, and decision-grade outputs.

Responsibility

This skill owns research workflow only:

• Scope: question framing, source collection, validation, synthesis, and reporting.
• Out of scope: direct production code changes unless explicitly requested.

Path Policy

• temp/: temporary downloads, extracted notes, throwaway validation scripts
• documents/drafts/: evolving research narrative and interim conclusions
• documents/reference/: stable references, papers, technical source summaries
• documents/final/: final Korean report for stakeholders

Protocol

1. Frame the research

• Define objective, constraints, and success criteria.
• Split into answerable sub-questions.

2. Collect evidence

• Prefer official docs, primary papers, and authoritative specs.
• Record source metadata early: title, link, version, date.

3. Validate claims

• Cross-check important claims with at least two independent sources.
• Flag unresolved contradictions explicitly.

4. Synthesize findings

• Build a concise argument chain from evidence to conclusion.
• Separate facts, assumptions, and recommendations.

5. Publish by stage

• Store working content in documents/drafts/.
• Save reusable source summaries in documents/reference/.
• Publish final stakeholder output to documents/final/ in Korean.

Output Contract

Every delivery includes:

• Research question
• Methods and source list
• Verified findings
• Risks and unknowns
• Actionable next steps

Quality Bar

• Claims are citation-backed
• Scope boundaries are explicit
• Conclusions are operationally useful, not generic