Academic Research Skills Tutorial: Full Research-to-Publication Pipeline with Claude Code

Created

Have you ever asked AI to help write a paper, only to find fabricated citations, logical contradictions, or prose that screams “machine-generated”? Academic Research Skills (ARS) solves this — not by writing your paper for you, but by turning AI into a research assistant that handles the grunt work: literature hunting, citation formatting, data verification, and logical consistency checks. You focus on what actually requires human judgment.

What is Academic Research Skills? ARS is an open-source Claude Code academic research skill suite (CC BY-NC 4.0 license) consisting of 4 core modules: Deep Research (13 agents, 7 modes), Academic Paper (12 agents, 10 modes), Academic Paper Reviewer (7 agents, 6 modes), and Academic Pipeline (10-stage orchestrator). The entire workflow emphasizes human-in-the-loop collaboration — AI handles the grunt work, humans handle decisions and judgment.

Prerequisites

  • AI Coding Assistant: Claude Code (latest version)
  • API Key: ANTHROPIC_API_KEY environment variable
  • Optional Dependencies:
    • Pandoc for DOCX output
    • tectonic + Source Han Serif TC for APA 7.0 PDF output

💡 Tip: If you use Codex CLI instead of Claude Code, install the sibling distribution academic-research-skills-codex — identical workflow content, Codex-native packaging.

Overview

ARS is designed around the principle of human-in-the-loop collaboration, not full automation. It draws insights from Lu et al. (2026, Nature 651:914-919) — The AI Scientist, the first fully autonomous AI research system to pass blind peer review, which identified critical failure modes (implementation bugs, hallucinated results, shortcut reliance) that prove pure automation unreliable. ARS addresses these through integrity verification gates at Stage 2.5 and Stage 4.5.

The system comprises 4 skill modules:

Module Agents Modes Core Function
Deep Research 13 7 Literature survey, systematic review, fact-checking
Academic Paper 12 10 Paper writing, style calibration, LaTeX output
Academic Paper Reviewer 7 6 Multi-perspective peer review, 0-100 scoring
Academic Pipeline — — 10-stage pipeline orchestration

Step 1: Install ARS

In Claude Code, run:

1
2
/plugin marketplace add Imbad0202/academic-research-skills
/plugin install academic-research-skills

Verify installation:

1
/ars-plan

This starts a Socratic dialogue to help you plan your paper structure.

1
2
3
git clone https://github.com/Imbad0202/academic-research-skills.git
cd academic-research-skills
ln -s "$(pwd)" ~/.claude/skills/academic-research-skills

Verify Installation

1
/ars-lit-review "your research topic"

If ARS begins a literature survey, installation is successful.

Step 2: Master the 4 Core Modules

1. Deep Research

A 13-agent research team with 7 modes:

1
2
3
4
5
6
7
"Research the impact of AI on higher education"       → full mode
"Give me a quick brief on X"                          → quick mode
"Do a systematic review on X with PRISMA"             → systematic-review mode
"Guide my research on X"                              → socratic mode (guided)
"Fact-check these claims"                             → fact-check mode
"Do a literature review on X"                         → lit-review mode
"Review this paper's research quality"                → review mode

💡 Tip: If you’re new to ARS, start with Socratic mode — it helps you clarify your research direction through questions rather than handing you a complete survey report.

2. Academic Paper

A 12-agent writing pipeline with 10 modes:

1
2
3
4
5
6
7
8
9
10
"Write a paper on X"                                  → full mode
"Guide me through writing a paper"                    → plan mode (guided)
"Build a paper outline"                               → outline-only mode
"I have a draft, here are reviewer comments"          → revision mode
"Parse these reviewer comments into a roadmap"        → revision-coach mode
"Write an abstract for this paper"                    → abstract-only mode
"Turn this into a literature review paper"            → lit-review mode
"Convert to LaTeX" / "Convert citations to IEEE"      → format-convert mode
"Check citations"                                     → citation-check mode
"Generate an AI disclosure statement for NeurIPS"     → disclosure mode

Key Features:

  • Style Calibration: Provide 3+ past papers and ARS learns your writing voice
  • Writing Quality Check: Auto-detects AI high-frequency terms (25 words), excessive dashes, structural patterns
  • LaTeX Output: Supports APA 7.0, IEEE, Chicago formats

3. Academic Paper Reviewer

A 7-agent multi-perspective review team with 6 modes:

1
2
3
4
5
6
"Review this paper"                                   → full mode (EIC + R1/R2/R3 + Devil's Advocate)
"Quick assessment of this paper"                      → quick mode
"Guide me to improve this paper"                      → guided mode
"Check the methodology"                               → methodology-focus mode
"Verify the revisions"                                → re-review mode
"Calibrate this reviewer against my gold set"         → calibration mode

Scoring: 0-100 scale. ≥80 Accept, 65-79 Minor Revision, 50-64 Major Revision, <50 Reject.

4. Academic Pipeline

A 10-stage full-pipeline orchestrator:

Stage Name Description
1 RESEARCH Literature survey, output research question brief
2 WRITE Paper writing, generate first draft
2.5 INTEGRITY GATE Citation verification, statistical checks
3 REVIEW Peer review (multi-perspective)
3’ RE-REVIEW Revision verification
4 REVISE Revise based on reviewer comments
4.5 INTEGRITY GATE Second integrity verification
5 FINALIZE Format, generate PDF
6 PROCESS SUMMARY Auto-generate collaboration quality report

Start the full pipeline:

1
"I want to write a complete research paper on [your topic]"

Step 3: Understand ARS’s Anti-Hallucination System

ARS’s core advantage lies in its anti-hallucination system:

Citation Verification

  • Semantic Scholar API Verification: Tier 0 programmatic reference existence check
  • Claim Verification Audit (v3.8): Set ARS_CLAIM_AUDIT=1 to verify that citations actually support the claims made in the text
  • Trust Chain Provenance: Every citation carries source provenance metadata

Integrity Verification Gates

  • Stage 2.5: First integrity check after draft completion — runs 7-mode AI research failure checklist
  • Stage 4.5: Second integrity check after revisions — confirms zero regressions

Anti-Sycophancy Mechanisms

  • Devil’s Advocate Concession Threshold: DA must score rebuttals 1-5 before responding; concession only at ≥4
  • No consecutive concessions allowed; concession rate tracked
  • Frame-lock detection: Checks each round for entrapment in human-set frames

Step 4: Supported Structures and Citation Formats

Paper Structures

  • IMRaD (empirical research)
  • Thematic Literature Review
  • Theoretical Analysis
  • Case Study
  • Policy Brief
  • Conference Paper

Citation Formats

  • APA 7.0 (default, with Chinese citation rules)
  • Chicago (Notes & Author-Date)
  • MLA
  • IEEE
  • Vancouver

Output Languages

  • Traditional Chinese: Default when user writes in Chinese
  • English: Default when user writes in English
  • Bilingual abstracts supported

FAQ

Q: How much does a full pipeline cost?
A: According to the official docs, a 15,000-word paper costs approximately $4-6 through the full pipeline. Recommended Claude Code settings: Skip Permissions enabled.

Q: Which AI assistants are supported?
A: Primarily Claude Code. For Codex CLI users, install the academic-research-skills-codex sibling distribution.

Q: Does ARS write my paper for me?
A: No. ARS follows the principle “AI is your copilot, not the pilot.” It handles literature hunting, formatting, and verification — you define the research question, choose the method, and interpret the data.

Q: How do I use it in Chinese?
A: ARS supports Traditional Chinese as the default output language. Socratic and Plan modes use intent-based activation that works in any language without modification.

Q: What is a Material Passport?
A: Material Passport is ARS’s cross-session resume mechanism, recording all key data from the paper creation process (research questions, methodology blueprint, citation lists, version history) for resuming from a new session.

Advanced Tips

Using the Experiment Agent

If your research involves running experiments (code or human studies), insert the Experiment Agent between ARS Stage 1 and Stage 2:

1
ARS Stage 1 RESEARCH → Experiment Agent → ARS Stage 2 WRITE

Cross-Model Verification

Enable cross-model integrity verification with GPT-5.4 Pro or Gemini 3.1 Pro:

1
export ARS_CROSS_MODEL=true

Import Existing Literature

Use Material Passport’s literature_corpus[] field to import your existing literature (PDF folders, Zotero exports, Obsidian notes):

1
2
3
4
5
literature_corpus:
  - authors: [...]
    year: 2025
    title: "Paper Title"
    source_pointer: "path/to/paper.pdf"

Conclusion

You’ve learned the core capabilities of ARS:

  1. ✅ Installed and configured Academic Research Skills
  2. ✅ Mastered the 4 core modules (Deep Research / Academic Paper / Reviewer / Pipeline)
  3. ✅ Understood the 10-stage paper pipeline workflow
  4. ✅ Learned the anti-hallucination system (citation verification, integrity gates, anti-sycophancy)
  5. ✅ Explored supported paper structures and citation formats

ARS’s core value is making AI a reliable research assistant rather than an automated paper generator that might fabricate citations and produce logical contradictions. Through human-in-the-loop collaboration, you focus on research that requires human judgment while AI handles the time-consuming grunt work.

📖 Official repo: github.com/Imbad0202/academic-research-skills
📖 Architecture: docs/ARCHITECTURE.md
📖 Setup Guide: docs/SETUP.md

How to cite this article: Based on ARS official README v3.9.4.2 (verified 2026-05-22). References Lu et al. (2026, Nature 651:914-919) and Zhao et al. (2026-05, arXiv:2605.07723) as academic background.