claude-code-academic-workflow

Skill by ara.so — Claude Code Skills collection.

What This Is

A production-ready template for AI-assisted academic work using Claude Code. Provides specialized agents for reviewing slides, papers, and data analysis; quality scoring with 80/90/95 thresholds; adversarial QA loops; and reproducibility verification. Originally extracted from a PhD course, now community-maintained with 1200+ GitHub stars.

Core capabilities:

• Contractor mode — describe what you want, Claude plans and orchestrates specialized agents
• 18+ focused agents — proofreader, slide-auditor, pedagogy-reviewer, r-reviewer, domain-referee, methods-referee, editor
• Quality gates — scores 0-100, auto-blocks commits below 80
• Adversarial QA — critic + fixer loop until approval or 5 rounds
• Verification layers — claim verification, reproducibility audit, AI-voice detection, multi-referee variance sampling