deep-research

name

deep-research

description

Deep Research (8-step methodology) — Transform vague topics into high-quality, deliverable research reports. Systematic fact extraction, source tiering (L1>L2>L3>L4), time-sensitivity assessment, and verifiable "Fact→Conclusion" chains. Use when: deep research, comprehensive report, thorough investigation, concept comparison, decision support, trend analysis. Inspired by wshuyi/deep-research + OpenAI Deep Research + HKUDS.

Deep Research Skill

8-step methodology: problem decomposition → source tiering → fact extraction → comparison framework → derivation chain → validation → final report.

Usage

python3 _scripts/deep_research.py "REST API vs GraphQL" --methodology --sources --debate --save
python3 _scripts/deep_research.py "What are the tradeoffs of AI agents?" --save

Options

• query — Research question or topic (required)
• --methodology — Full 8-step flow with intermediate artifacts saved to Sources/Deep Research/<topic>/
• --sources — Include external source recommendations
• --debate — Include steelmanned opposing views
• --top-k N — Max notes to retrieve (default: 15)
• --depth N — Graph expansion depth (default: 2)
• --save — Save final report to vault

Source Tiering

L1 (official docs/papers) > L2 (official blogs) > L3 (media/tutorials) > L4 (community/forums). Conclusions must trace to L1/L2.

Output

Sources/Deep Research - YYYY-MM-DD.md (or full artifact folder with --methodology)