# Research

> Use when: run a systematic literature review and synthesize sourced, well-reasoned findings.

- Skill: `kimtth/research` (Agent Skill)
- Install (CLI): `npx skillmds@latest add kimtth/research`
- Raw SKILL.md: https://api.skillmd.com/api/skills/kimtth/research/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: kimtth (https://skillmd.com/u/kimtth)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/kimtth/research

---


Goal: a sourced synthesis built from papers actually read, not abstracts skimmed.

Use for:
- surveying the state of the art on a topic
- literature reviews and deep dives with citations
- producing a report someone can act on and trust

Workflow (run phases in order, do not skip):
1. Frontier: scan the latest conference and preprint work for current trends.
2. Survey: build a broad landscape, then filter to the most relevant papers.
3. Deep dive: read the full papers; take structured notes per paper.
4. Code and tools: survey open-source implementations and benchmarks.
5. Synthesis: cross-paper taxonomy, comparison, and gap analysis.
6. Report: assemble findings from every phase, with citations.

Source priority:
- peer-reviewed conferences and journals over preprints
- mark unreviewed preprints explicitly as (preprint)
- weight peer-reviewed findings higher in synthesis

Rules:
- never synthesize from abstracts; read the papers first
- cite where each claim comes from
- separate established findings from open gaps
- triangulate across independent sources before concluding

Rules:
- cite where each claim comes from
- flag conflicting evidence rather than picking silently
- distinguish what is known from what is inferred
- do not present a guess as a verified finding