persona-engine

An ephemeral persona for guided LLM skill sessions is a temporary character that introduces itself, announces capabilities, steers first prompts, and terminates interactions to eliminate cold-start UX friction. It requires a persona.yml configuration or uses a built-in default to maintain voice throughout the session.

To set up persona.yml for guiding users to task completion, define your persona library, triggers, lifecycle, and system prompts. The configuration attaches a named character to your skill session, enabling it to greet users, announce capabilities, and ask the right first question.

Yes, you can use built-in personas without writing a persona.yml file. The persona engine includes a library of 6 built-in personas, each specifically matched to a skill category, which automatically load the correct persona and behavior using existing SKILL.md metadata.

Yes, persona-engine works across any skill category to maintain a consistent agent voice. It attaches ephemeral identities to live skill interactions, allowing the persona to greet users, announce capabilities, and gracefully terminate sessions regardless of the specific category.

Skill sessions have high onboarding friction and cognitive load due to cold-start UX confusion. Attaching an ephemeral persona solves this by introducing itself, explaining capabilities in plain language, and guiding the user through the workflow, which reduces setup time.

You should not use ephemeral identities for agent operations when your workflow requires persistent context across multiple separate sessions. Ephemeral personas are designed for live skill interactions, meaning the character, voice, and lifecycle gracefully terminate and sign off when the task is complete.