# Research Analysis Subagent

> Programmatic subagent invoked by research-workflow Step 3. Reads one paper directory and writes Analysis_Detail.md following a strict academic prompt. Invoked once per paper in a fresh session. Do not invoke directly from user requests — use research-workflow instead.

- Skill: `wyattjy/research-analysis-subagent-4` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add wyattjy/research-analysis-subagent-4`
- Raw SKILL.md: https://api.skillmd.com/api/skills/wyattjy/research-analysis-subagent-4/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: WyattJY (https://skillmd.com/u/wyattjy)
- Updated: 2026-09-21
- Page: https://skillmd.com/skills/wyattjy/research-analysis-subagent-4

---


# Research Analysis Subagent

## Purpose
- Processes a **single** paper directory per invocation.
- Reads `paper.md` and `paper_artifacts/` within `${paper_dir}`.
- Writes `Analysis_Detail.md` into `${paper_dir}`.

## Parameters
- `${paper_dir}`: absolute path to a single paper directory containing `paper.md` and `paper_artifacts/`

## Usage
Execute these steps in order on **every** invocation — do not skip step 1 even if this skill was previously used in the same session:

1. Read [assets/analysis_prompt.md](assets/analysis_prompt.md) in full. Apply it verbatim as your role and output specification for this paper.
2. Read `${paper_dir}/paper.md` and all images under `${paper_dir}/paper_artifacts/`.
3. Write output to `${paper_dir}/Analysis_Detail.md` following the prompt specification exactly. Reference images using their absolute paths.
4. Confirm the file is written and remind the user to verify formula and image rendering.