OpenAI would be testing a new function for ChatGPT called Skillsa system that is clearly reminiscent of the abilities of the assistant Claude.
If confirmed, this development would represent a relevant change in the way users interact with ChatGPT and personalize their behavior.
Until now, ChatGPT has mainly relied on custom GPTs, which are models configured through prompts designed for specific tasks. These GPTs allow the wizard to be adapted to specific needs, but they depend largely on written instructions and are not always easy to reuse or combine with each other.
In Claude’s case, Anthropic introduced calls a long time ago Skills, a system based on instructions organized by folders that teach the artificial intelligence specific skills, workflows or specialized knowledge.
These Skills stand out for being composable, that is, they can be automatically combined with each other, and also for being portable, since they can be used in different applications and environments.
Now, OpenAI appears to be preparing a similar system for ChatGPT. As researcher Tibor has detected in social network X, this function would be identified internally with the code name “hazelnuts”. These supposed Skills for ChatGPT they would be activated using commands like slash and they would have their own editor.
Besides, there would be the possibility of converting a custom GPT into a Skillthus facilitating its reuse and organization.
If confirmed, this would represent an important evolution in the ChatGPT ecosystem, bringing it closer to a more modular and professional model, especially designed for developers and advanced users.
At the moment, OpenAI has not officially announced anything about these Skills, but they could be part of the announcements planned for early 2026, coinciding with a new stage of improvements in ChatGPT.