s
sere.loyalty_2s
Personas are useful for setting the agent’s general behavior, but reusable workflows need something more modular.
Currently, I have to fit both my general research standards and detailed workflows into a single persona instructions field with a 10,000-character limit. Combining everything makes the prompt harder to maintain. Splitting it across personas means duplicating research rules and keeping those copies consistent.
I would love support for reusable skills that work alongside persona instructions. The persona would define the agent’s general standards, while a skill would define how to carry out a particular task.
Requested behavior:
  • Let users create named skills with instructions, activation conditions, workflow steps, and approval requirements. Users should be able to invoke a skill explicitly, and the agent could suggest a relevant skill based on the conversation.
  • Agent can load the selected skill’s complete instructions reliably. Keep the workflow active across conversation turns, including temporary discussion or research detours.
  • Let the agent help users draft, test, and improve skills through conversation. It should propose changes for approval before saving them, ideally with version history and the ability to restore an earlier version.
The goal is to customize repeatable ways of working with the agent, not just its tone or response format, without cramming every workflow into one increasingly complicated prompt.