Feature Request: Auto-import skills and agents from user's .github repo

[sandeeprawat]

Summary

Allow Copilot CLI to automatically discover and load skills and agents from a user's <username>/.github repository on GitHub, similar to how .github repos already serve as a convention for profile READMEs and org-wide configurations.

Problem

Today, personal skills and agents must be manually placed in ~/.copilot/skills/ and ~/.copilot/agents/. This means:

• Skills/agents aren't synced across machines — if I switch between a laptop, desktop, and codespace, I have to manually copy files to each environment.
• There's no easy way to share a curated set of skills/agents with teammates ("just clone my .github repo").
• New machine setup requires extra steps beyond just installing Copilot CLI.

Proposed Solution

Option A — Automatic sync:
If the authenticated user has a .github repository (e.g. sandeeprawat/.github), Copilot CLI automatically fetches (or git-clones/pulls) it and loads:

• skills/ → as personal skills
• agents/ → as personal agents

This would happen at CLI startup (with caching so it doesn't slow things down on every launch). A config flag could control the behavior: "importGitHubHomeRepo": true | false.

Option B — Manual command:
A new slash command like /skills import-github or /import-dotgithub that fetches the user's .github repo on demand and registers the skills/agents found in it.

Ideally both — auto-sync by default with a manual command for on-demand refresh.

Benefits

• Zero-config portability — Skills and agents follow the user across machines, codespaces, and environments automatically.
• Leverages existing GitHub convention — The <user>/.github repo is already a well-known pattern.
• Shareability — Users can point colleagues to their .github repo as a curated skill/agent collection.
• Org-level extension — Could naturally extend to org .github repos (.github-private) for org-wide skills, complementing the existing org-level agent support.

Open Questions

• Should this be opt-in or opt-out?
• How to handle conflicts between repo-fetched skills and locally defined skills with the same name?
• Caching/refresh strategy — pull on every startup vs. periodic vs. manual only?
• Security: should there be a trust prompt before loading remotely-fetched skills for the first time?