> TL;DR: An MIT-licensed set of Claude Code Skills covering the full academic research lifecycle — 13-agent deep research → 12-agent paper writing → 7-agent peer review → a 10-stage automated pipeline. A 15,000-word paper costs roughly $4-6 in API fees plus 2-4 hours of collaboration.
Why did this repo gain 11,600 stars in one week?
In May 2026, academic-research-skills topped GitHub Trending with 20,268 total stars and 11,600 new stars in 7 days. It reflects the community's deep understanding of Claude Code's "Skill" model — not letting AI do your research for you, but letting AI handle the mechanical work while you focus on the parts that genuinely require your expertise.
> "AI is your copilot, not the pilot."
Core Skills Overview
1. Deep Research (v2.9.4) — a 13-agent research team
| Mode | Use case | Depth |
|------|----------|-------|
| full | Complete research | Balanced |
| quick | 30-minute rapid briefing | High fidelity |
| systematic-review | PRISMA-compliant systematic review | High fidelity |
| socratic | Guided research dialogue | Originality |
| fact-check | Fact checking | High fidelity |
| lit-review | Literature review | High fidelity |
| review | Paper evaluation | Balanced |
Core capabilities:
- Socratic guided mode — helps you clarify vague ideas
- PRISMA systematic reviews + meta-analysis
- Intent detection — automatically identifies what you actually want to research
- Conversation health monitoring — prevents the AI from drifting off track
- Semantic Scholar API verification — eliminates hallucinated citations
- Optional cross-model adversarial verification
- Style Calibration — provide 3+ of your past papers and the system learns your writing voice
- Writing Quality Check — warnings for 25 common AI-frequent words, dash overuse, throat-clearing openings, structural patterns, and burstiness
- LaTeX Hardening — APA 7.0
apa7class / IEEE / Chicago - VLM Figure Verification — validates figures and charts
- Anti-Leakage Protocol — prevents information leakage
- Read-Only Constraint — reviewer agents cannot modify the manuscript, only suggest changes (mirroring the boundaries of real peer review)
- R&R Traceability Matrix — tracks whether each reviewer comment was addressed by the author
- 3 cognitive framework files:
- Toulmin argumentation model + Bradford Hill causal reasoning
- Three-lens review (internal validity / external validity / contribution)
- Research gap definition (real gap vs. rhetorical gap)
2. Academic Paper (v3.1.2) — 12-agent paper writing
| Mode | Use case |
|------|----------|
| full | Complete paper |
| plan | Guided writing |
| outline-only | Outline only |
| revision | Revise per reviewer comments |
| revision-coach | Reviewer comment analysis |
| abstract-only | Abstract only |
| lit-review | Literature-review paper |
| format-convert | LaTeX / citation style conversion |
| citation-check | Citation checking |
| disclosure | AI disclosure statement (NeurIPS, etc.) |
Core capabilities:
3. Academic Paper Reviewer (v1.9.1) — 7-agent multi-perspective review
| Role | Responsibility | |------|----------------| | EIC (Editor-in-Chief) | Overall quality gatekeeping | | R1 / R2 / R3 | Three dynamic reviewers, each emphasizing different dimensions | | Devil's Advocate | Challenges the paper across 8 dimensions | | Quality Rubric | 0-100 score: ≥80 Accept, 65-79 Minor, 50-64 Major, <50 Reject |
Key design decisions: