# Research Code Style

> Initialize empty Python research repositories or review populated Python research repositories for minimal style, importable src modules, run_scripts shell launchers, uv-based commands, and clean experiment folders. Use for new empty research repos, AGENTS.md coding rules, or repository-structure reviews. For non-empty repos, review only; do not directly modify, move, delete, or reorganize files.

- Skill: `chendaiwei-99/research-code-style` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add chendaiwei-99/research-code-style`
- Raw SKILL.md: https://api.skillmd.com/api/skills/chendaiwei-99/research-code-style/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Research & Search
- Author: ChenDaiwei-99 (https://skillmd.com/u/chendaiwei-99)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/chendaiwei-99/research-code-style

---


# Research Code Style

Use this skill to initialize empty research repos or review populated ones against a minimal Python research-code standard.

## Workflow

1. Decide the mode:
   - Empty repo: initialize a minimal layout.
   - Populated repo: review only; do not edit files.
2. Read local context first: `AGENTS.md`, `pyproject.toml`, README files, and existing module patterns.
3. Apply [rule-of-thumbs.md](references/rule-of-thumbs.md).
4. Do not hand-create `pyproject.toml` or `uv.lock`; the user initializes them with `uv init`.
5. Do not run expensive training or experiment commands unless the user explicitly asks.

## Output

- Empty repo: create only the minimal source/layout skeleton.
- Populated repo: return findings with paths, risks, and suggested moves.
- AGENTS request: return a concise `AGENTS.md` section from the reference.