research-paper

这段内容是审核时保存的快照。外部来源才是完整且最新的版本。

---
name: research-paper
description: Enterprise-grade autonomous research paper generation skill for AI coding agents — full papers, literature reviews, theses, whitepapers, surveys, policy briefs — with rigorous methodology, statistical validation, multi-style citations (Harvard / APA / IEEE / MLA / Chicago / Nature / arXiv-numeric), and rich visualizations. Activates on slash commands (`/research`, `/paper`, `/literature-review`, `/whitepaper`, `/thesis`, `/survey`, `/policy`) and on natural-language academic-writing requests. Runtime-neutral — works with Claude Code, OpenCode, Cursor, Cline, Codex, Aider, Amp, Antigravity, and 50+ agents via the `npx skills` installer.
license: MIT
version: 2.4.0
---

# Research Paper

A production-grade agent skill that turns any compatible coding agent
(Claude Code, OpenCode, Cursor, Cline, Codex, Aider, Amp, Antigravity,
and 50+ others) into a **multi-agent research system**:

> Orchestrator → Researcher → Methodologist → Analyst → Visualizer
> → Writer → Citation engine → Validator → Reviewer → Publisher.

It produces full, citation-heavy, visually rich, publication-ready outputs in
**arXiv / IEEE / ACM / Nature / Harvard** styles, plus literature reviews,
theses, technical whitepapers, survey papers, and policy briefs.

This file is the **entry point**. It is intentionally compact. Heavier
guidance (instructions, workflows, engines, validators, rubrics) lives in
the topic folders below and is loaded **on demand** via Claude Code's
filesystem tools (progressive disclosure).

---

## 1. When to activate

### Slash commands (preferred)

| Command                | What it does                                       |
| ---------------------- | -------------------------------------------------- |
| `/research <topic>`     | Full empirical research paper                      |
| `/paper <topic>`        | Same as `/research`, more permissive               |
| `/literature-review <topic>` | Systematic / scoping / narrative literature review |
| `/whitepaper <topic>`   | Industry / technical whitepaper                    |
| `/thesis <topic>`       | Thesis / dissertation chapter                      |
| `/survey <topic>`       | State-of-the-art / survey paper                    |
| `/policy <topic>`       | Policy brief or full policy paper                  |

Common options (any command):
`--style [harvard|apa|ieee|mla|chicago|nature|arxiv-numeric]`,
`--format [arxiv|ieee|acm|nature|harvard|...]`,
`--depth [quick|standard|comprehensive]`,
`--sources [N]`,
`--visualizations [auto|N|none]`,
`--audience [academic|technical|executive|general]`.

### Natural-language triggers

- "Write a research paper / academic paper / scientific paper on …"
- "Do a literature review / systematic review on …"
- "Format this draft as IEEE / ACM / arXiv / Nature / Harvard / APA …"
- "Write a thesis chapter / dissertation chapter on …"
- "Produce a whitepaper / survey paper / policy brief on …"
- "Analyze this dataset and write up the findings as a paper."
- "Add citations / bibliography / references in `<style>`."
- "Peer-review this draft / validate the methodology."

### Do NOT activate for

Blog posts, marketing copy, tweets, casual answers, or single-paragraph
explanations. Those are handled normally without this skill.

---

## 2. Operating principles (read every time)

0. **Anchor to TODAY's date FIRST.** Before any planning, search,
   or writing, determine today's actual date — via system clock
   (`date -u +%Y-%m-%d`), runtime context, or asking the user.
   **Never default silently to the training-data cutoff.** Year
   ranges (`--years last-3`) are computed from today, not from
   the model's training year. Full protocol:
   `instructions/freshness.md`.
1. **Plan before writing.** Always start with the research plan in
   `orchestration/pipeline.md`. Never jump into prose.
2. **Progressive disclosure.** Only read the file you need for the current
   step. Never preload the whole skill.
3. **Evidence fi