Agent Skills turns Codex workflows reusable

This is less a prompt library than a packaging format for operational know-how, which is why the repo is landing so well with developers. OpenAI is nudging agent work from ad hoc chat tricks toward reusable, auditable workflows.

• –Skills use progressive disclosure, loading metadata first and full instructions only when needed, which helps agents manage context more efficiently.
• –The system, curated, and experimental split creates a practical path from built-in defaults to team-specific workflow distribution.
• –The install flow through $skill-installer makes skills feel like portable extensions instead of copy-pasted prompts.
• –Optional agents/openai.yaml metadata adds invocation policy and tool dependencies, pointing toward a richer ecosystem of interoperable agent capabilities.
• –The repo’s rapid GitHub traction suggests strong demand for shareable agent workflows that are inspectable in plain files.