What it does

This repo packages the Academic Research Skills (ARS) suite—previously a Claude Code plugin—as a Codex-native skill. You install it with a Python script into ~/.codex/skills/, then invoke $academic-research-suite to route your request through one of five workflows: deep-research, paper drafting, peer review, full pipeline, or experiment planning. The router picks the mode based on your prompt, so “I have notes but no outline” and “review this manuscript for desk-reject risks” go to different subsystems.

The interesting bit

The packaging is deliberately thin: upstream ARS content is vendored unmodified and adapted through a Codex router in SKILL.md. Claude-specific hooks become traceability metadata; slash commands become plain-text aliases (ars-plan, ars-reviewer); and cross-model verification, if you want it, explicitly calls Claude Opus 4.7 via Anthropic’s API—not OpenAI’s. It’s a compatibility layer that knows exactly what it can’t port.

Key highlights

• Five workflows under one skill: research scoping, paper writing, peer review, end-to-end pipeline, and experiment design
• Socratic routing: vague topics get narrowing questions before any drafting happens
• Claude-style aliases (ars-plan, ars-full) emulated inside the single Codex skill
• Optional cross-model verification via explicit ARS_CROSS_MODEL=claude-opus-4.7 + ANTHROPIC_API_KEY
• Version 0.1.9 tracks upstream ARS v3.10.0; package version is independent of vendored content

Caveats

• Codex does not register native slash commands; aliases only work if your client passes them through as user text
• Cross-model verification is disabled by default and requires a separate Anthropic API key
• Some upstream Claude features (plugin loader, .github/ CI gates, background agent spawning) are excluded or require explicit user delegation

Verdict

Researchers already using Codex who want structured, integrity-gated academic workflows should try this. If you’re happy with Claude Code, the original ARS repo is the better-supported path; this is primarily for Codex-native environments.