Enhancement: Allow Adding External Skill Repositories in Codex App

Resolved 💬 3 comments Opened Mar 1, 2026 by aero-oli Closed Apr 11, 2026
💡 Likely answer: A maintainer (github-actions[bot], contributor) responded on this thread — see the highlighted reply below.

What variant of Codex are you using?

App

What feature would you like to see?

🚀 Enhancement: Allow Adding External Skill Repositories in Codex App

Summary

Add support in the Codex App → Skills tab to register external Git repositories that contain skills, so their skills appear in the “Available to Install” list — alongside OpenAI-provided skills.

Problem

Currently, the Skills tab only shows skills published/curated by OpenAI.

There is no way to:

  • Add a third-party skills repository
  • Browse skills from that repository inside the app
  • Install those skills directly via the UI

This limits discoverability and prevents community-driven skill ecosystems from emerging.

Proposed Enhancement

In the Skills tab, add functionality to:

  1. Register external Git repositories
  • Public GitHub repos (e.g. https://github.com/K-Dense-AI/claude-scientific-skills)
  • Potentially support:
  • HTTPS
  • SSH
  • Local filesystem paths
  1. Automatically index skills in that repo
  • Detect skill definitions (e.g. via manifest/spec files)
  • Validate structure
  • Display each skill in the installable list
  1. Display source metadata
  • Show which repository the skill originates from
  • Display version/tag/commit
  • Indicate trust level (e.g. “External” vs “OpenAI Official”)
  1. Install directly from UI
  • Clone or fetch repo
  • Install selected skill
  • Pin version or track branch

---

Example Use Case

I add:

https://github.com/K-Dense-AI/claude-scientific-skills

The app:

  • Scans the repository
  • Detects all valid skills inside it
  • Displays them in the Skills tab under a new section like:
External Repositories
  claude-scientific-skills (GitHub)

Each skill in that repo then appears as installable just like official skills.

---

Why This Matters

  • Enables a community skill ecosystem
  • Encourages reusable skill libraries
  • Removes friction from sharing skills
  • Aligns with MCP’s modular philosophy
  • Makes Codex feel extensible rather than closed

Right now, skill reuse requires manual copying or configuration workarounds. This feature would make skill distribution first-class.

---

Optional Enhancements

  • Repo trust warnings
  • Skill signing / verification
  • Auto-update support
  • Private repo support (via token)
  • Organization-wide skill registries
  • Allow pinning to tag/commit

---

UX Suggestion

Add in the Skills tab:

  • “➕ Add Skill Repository” button
  • Repository management panel
  • Toggle to enable/disable a repo
  • Refresh button
  • Clear separation between:
  • Official Skills
  • External Skills

---

Technical Considerations

  • Define canonical skill discovery pattern (e.g. /skills/*/skill.yaml)
  • Version compatibility validation
  • Safe sandboxing for untrusted skill repos
  • Clear metadata schema

Additional information

_No response_

View original on GitHub ↗

3 Comments

github-actions[bot] contributor · 5 months ago

Potential duplicates detected. Please review them and close your issue if it is a duplicate.

  • #13074

Powered by Codex Action

etraut-openai contributor · 4 months ago

This feature request hasn't received enough upvotes, so I'm going to close it.

Tomchen070814 · 1 month ago

Hi @aero-oli — your proposal describes almost exactly the workflow I have been prototyping around.

I’m building Bitcase, a free public Alpha for collecting external GitHub Skill repositories, inspecting SKILL.md before saving, preserving source and trust information, avoiding duplicates, and assembling a minimal Skill Stack for a real Codex project.

Important limitation: Bitcase is not yet a native Codex installer, so it does not add repositories directly inside the Codex App. I’m currently validating whether the discovery, inspection, source management, and project-matching layer is genuinely useful before building deeper installation and synchronization features.

Would you be willing to test it with K-Dense-AI/claude-scientific-skills or another external repository you already use?

https://bitcase-skill-library.tomcjq070814.chatgpt.site/?utm_source=github&utm_campaign=alpha_research&utm_content=aero_oli_13174

No account is required. I’m looking for critical feedback rather than promotion—especially what is missing compared with the workflow you described here.