Love the pattern of treating workflows as a library instead of constantly re-pasting runbooks. The distinction between Skills (global habits) and AGENTS.md (repo-specific docs) is super clean. The compact Skills section approach keeps token costs down while still giving Codex visibility. Curious how well this scales when you have 20+ skills across a team.
Love the pattern of treating workflows as a library instead of constantly re-pasting runbooks. The distinction between Skills (global habits) and AGENTS.md (repo-specific docs) is super clean. The compact Skills section approach keeps token costs down while still giving Codex visibility. Curious how well this scales when you have 20+ skills across a team.
The implementation currently is very similar to MCP server discovery. That's probably what we can lean on when we look at scalability.