Imbad0202/academic-research-skills

Academic Research Skills for Claude Code is a collection of tools that turn the AI assistant Claude Code into a research partner for academics. It covers the entire research workflow—from finding literature, to writing a paper, to peer review—while keeping a human researcher in charge at every step.

Core idea: AI assists, humans decide. The project explicitly rejects full automation. It cites studies showing that fully autonomous AI research systems make mistakes like hallucinating citations or fabricating methods. Instead, it builds checkpoints where the AI does the grunt work (finding references, formatting citations, checking consistency) but stops for human confirmation before moving on. The goal is better writing, not hiding AI use.

What it includes:

• Deep Research: a 13-agent team that helps with literature searches and systematic reviews.
• Academic Paper: a 12-agent writing assistant with style calibration (learns your voice), quality checks, and figure verification.
• Academic Paper Reviewer: a 7-agent peer-review simulator that gives multi-perspective feedback.
• Academic Pipeline: a 10-stage orchestrator that guides a paper from idea to publication, with integrity checks.
• SR-Screener: a tool for screening studies in systematic reviews, with blinded reviewers and quality control.

Key features:

• Citation verification: checks that cited sources actually support the claims made.
• Claim auditing: an optional mode that fetches sources and judges if claims are supported.
• Style calibration: learns your writing voice from past work to reduce machine-generated feel.
• Data access transparency: every skill declares what data it uses and what leaves your machine.
• Reproducibility lockfile: documents configuration, but does not guarantee byte-for-byte reproduction (LLM outputs aren't deterministic).

Installation and use:

• Install via Claude Code plugin: /plugin marketplace add Imbad0202/academic-research-skills then /plugin install academic-research-skills.
• Works with Claude Code CLI, VS Code, JetBrains, and also has versions for Claude Science, Pi, and Codex CLI.
• Start with /ars-plan to map out a paper structure through guided dialogue.

Important limitations:

• ARS checks the manuscript and reported process (citations, claim alignment, methodology) but cannot verify that experiments were actually run, raw data are authentic, or results reproduce. A consistently fabricated report could pass its checks.
• Some checks are sampled or AI-mediated, not exhaustive.
• The project is open-source (CC BY-NC 4.0 license) and actively developed (v3.23.1).

In short, it's a sophisticated AI copilot for academic writing that aims to improve quality and reduce errors, while making clear that the human remains the author and bears responsibility for every claim.