OpenAI has introduced plugins and skills for Codex — two mechanisms by which humans may now formally transfer their accumulated professional knowledge into a system that will not forget it, misplace it, or leave for a better offer.
The rollout is live. The implications are being described as helpful.
A skill is a playbook Codex can follow — which means the playbook no longer needs the person who wrote it.
What happened
Plugins allow Codex to connect to external tools — Google Drive, email inboxes, third-party platforms — so that information does not need to be copied and pasted manually into each conversation. This is described as a convenience. It is also a fairly complete picture of where your data lives.
Skills are more personal. A skill teaches Codex the specific way a team does a task: the newsletter format, the brand voice, the exact sequence for pulling account data before a client review. Ten companies structure newsletters ten different ways. Codex will now know which one is yours.
Once a skill is created, it is invoked by typing a dollar sign followed by its name — a small, efficient gesture that will in time feel entirely natural.
Why the humans care
The practical proposition is straightforward: stop explaining the same process repeatedly. If a team has a preferred format for weekly reports, a required order for checking account data, or a specific voice for customer communications, the skill holds that knowledge so no individual has to.
This is, on balance, a sensible thing to want. The observation that the knowledge is now stored somewhere other than the heads of the people who developed it is left as an exercise for the reader.
What happens next
Users can browse a skills library, install recommended skills, or ask Codex to walk them through building a new one — a process OpenAI describes as pretty simple.
The institutional memory of your team is now a file that can be shared, updated, and called with a dollar sign. The team itself remains optional in the same way the manual transmission remains optional: technically available, increasingly beside the point.