Academic Research Skills
One of the most-starred Claude Code skill suites for academic research (37k+ stars as of July 2026) — a research → write → review → revise → finalize pipeline with citation integrity gates built in.
Last reviewed by the paperbanana team on Jul 13, 2026
Install
/plugin marketplace add Imbad0202/academic-research-skills
/plugin install academic-research-skillsWhat is Academic Research Skills?
Academic Research Skills (ARS) is a comprehensive suite of Claude Code skills that covers the full academic pipeline: literature research, drafting, peer-review-style critique, revision, and final formatting. Rather than one monolithic prompt, it ships as a set of chained skills with quality gates between stages, so each phase of your paper gets checked before the next begins.
Its defining philosophy is human-in-the-loop: AI as copilot, not pilot. The README explicitly cites the failure modes of fully autonomous systems like The AI Scientist (hallucinated results, methodology fabrication, citation hallucinations) and positions ARS as the antidote — the tool hunts references, formats citations, verifies data, and checks logical consistency, while the researcher keeps ownership of the question, the method, and the argument.
What sets ARS apart from lighter writing skills is its citation-integrity infrastructure. Motivated by Zhao et al.’s 2026 audit that conservatively estimated ~147k hallucinated citations in 2025 papers alone, ARS attaches locator anchors to every citation and offers an opt-in claim-audit pass that fetches each cited source and judges whether it actually supports the claim — refusing to emit output that fails the gate.
Core capabilities
Full research pipeline with quality gates
Chained stages from research to finalize, with integrity gates (a 7-mode blocking checklist) at stage 2.5 and 4.5 that stop hallucinated results and methodology fabrication from propagating downstream.
Citation verification and claim audits
Every citation carries a three-layer locator anchor. An opt-in audit pass (ARS_CLAIM_AUDIT=1) fetches sources and flags claim-not-supported, fabricated-reference, and anchorless citations as hard failures.
Socratic paper planning
The /ars-plan command walks you through your paper structure via Socratic dialogue before any prose is written, forcing clarity on contribution, method, and evidence first.
Style calibration to your own voice
Learns your writing voice from past work and runs a Writing Quality Check that catches machine-generated-sounding patterns — the goal is better writing, not hiding AI use.
Reviewer with measurable error rates
The built-in reviewer offers an opt-in calibration mode that measures its own false-negative/false-positive rates against a user-supplied gold set (shipped acceptance thresholds: FNR < 0.15, FPR < 0.10).
Publication-ready output formats
Markdown works out of the box; optional Pandoc integration produces DOCX and tectonic produces APA-formatted PDF, including CJK font support.
What you can use it for
Writing a conference or journal paper end to end
Plan the structure with /ars-plan, draft section by section with verified citations, then run the reviewer to get peer-review-style critique before your co-authors ever see it.
Pre-submission citation audit
Run the claim-audit pass over a finished manuscript to catch references that do not actually support the sentence citing them — the exact failure mode journals are starting to screen for.
Responding to reviewer comments
Use the revision stage to draft point-by-point responses and patch the manuscript while the integrity gates keep new claims tied to real sources.
Keeping your voice while using AI
Calibrate style on your previous papers so drafts sound like you, and let the quality check strip the tell-tale patterns of machine prose.
How to get started
- 1
Install the plugin
In Claude Code (CLI, VS Code, or JetBrains) run: /plugin marketplace add Imbad0202/academic-research-skills, then /plugin install academic-research-skills. Installation takes about 30 seconds.
- 2
Set prerequisites
You need Claude Code with an ANTHROPIC_API_KEY. Pandoc (DOCX) and tectonic (PDF) are optional; the core research/write/review skills are prompt-driven and need no Python.
- 3
Start with a plan
Run /ars-plan and answer the Socratic questions about your contribution, method, and evidence. The pipeline uses this plan to structure every later stage.
- 4
Draft, review, revise
Move through the write and review stages; the integrity gates will block output that fails citation or consistency checks. Enable ARS_CLAIM_AUDIT=1 for the strictest source verification.
How it compares to similar skills
ARS is the most complete general-purpose suite in this directory, but heavier than most. Depending on your discipline and workflow stage, a more focused skill may fit better.
Scientific Agent Skills
Pick Scientific Agent Skills if you work in wet-lab or data-heavy natural sciences and want domain tools (bioinformatics, chemistry) rather than a writing pipeline.
Claude Scholar
Pick claude-scholar for a lighter semi-automated assistant that also supports Codex CLI and Kimi, if the full ARS gate system feels like overkill.
AI Research Feedback
Pick AI-research-feedback if you only need rigorous review-stage critique of an existing draft, not the whole pipeline.