Academic Research Skills for Codex (ARS-Codex)

  • Author/Source: Edward Cheng-I Wu, GitHub
  • Original: https://github.com/Imbad0202/academic-research-skills-codex

  • Key Ideas

  • Codex-native sibling distribution of Academic Research Skills. Same workflow content; repackaged as a single Codex skill $academic-research-suite with ars-* aliases.
  • Vendors the upstream ARS suite verbatim under skills/academic-research-suite/ars/, plus a Codex router in SKILL.md and OpenAI agents YAML at agents/openai.yaml.
  • Codex package version 0.1.6 tracks independently of the vendored ARS version (currently pinned to a specific upstream commit recorded in manifest.source_repositories[]).
  • Single-line install via the Codex skill installer: python ~/.codex/skills/.system/skill-installer/scripts/install-skill-from-github.py --repo Imbad0202/academic-research-skills-codex --ref main --path skills/academic-research-suite --method git.

  • Summary

Codex packaging of the ARS pipeline. The vendored content is identical to the Claude Code version; the value of the separate repo is purely distribution: Codex's skill system uses a single-suite layout rather than the multi-skill directory layout that Claude Code expects, and a single Codex skill router replaces Claude Code's /plugin marketplace flow. Users who prefer Codex CLI / Codex desktop for academic writing get the same agent rosters, integrity gates, anti-sycophancy protocols, and journal references without juggling two toolchains.

  • Relevance to Economics Research

A reminder that the Codex ecosystem has reached feature parity with Claude Code for skill-based workflows. For economists committed to the OpenAI stack (e.g., GPT-5.3-Codex users following Panjwani's course), this is the canonical port of the most-elaborate academic-research skill bundle. The cross-platform port also documents an interesting pattern: when the Claude Code and Codex skill formats diverge, vendoring the upstream and writing a thin adapter is more sustainable than dual-maintaining the workflow content.