Persona Distiller

Extract structured, evidence-backed AI persona profiles from historical chat logs, emails, and documents.

Install
cmdop skills install agensi-persona-distiller

The Agensi Persona Distiller skill processes raw historical data, including email threads, chat logs, and various documents, to generate structured AI persona specifications. It distinguishes between stable characteristics and temporary context to capture the authentic essence of a subject’s decision-making and communication style. This process moves beyond simple mimicry by systematically separating stable traits from temporary context, ensuring the captured essence is based on evidence and without including noise or private data. It achieves this by identifying repeated patterns in reasoning, communication posture, and boundary-setting. The output is a high-density English persona profile, regardless of the source language, containing specific sections for Decision Posture, Reasoning Style, and Interaction Patterns. Crucially, each profile includes a Confidence score and an Open Gaps section, which clearly indicates where the data was insufficient to make a strong inference. The skill supports evidence from multi-turn conversation archives such as Slack, Discord, and WhatsApp, professional email corpora, internal memos, authored articles, whitepapers, and technical documentation. This allows agents to analyze extensive communication histories to build robust persona models.

Use cases

  • Use it to create consistent digital twin AI agents.
  • Use it to develop specialized customer support AI agents.
  • Use it to build automated executive assistant AI agents that maintain a professional voice.
  • Use it to analyze communication patterns from chat logs.
  • Use it to extract reasoning styles from professional emails.
  • Use it to define interaction patterns from technical documentation.

When to use it

  • When an AI agent needs to maintain a consistent professional voice.
  • When creating AI agents based on an individual's historical communication.
  • When a rigorous analytical framework for persona extraction is required.
  • When source data includes multi-turn conversations, emails, or documents.
  • When requiring a confidence score and identified data gaps for persona profiles.

When not to use it

  • When the goal is simple AI mimicry without detailed analysis.
  • When no historical data is available for persona extraction.
  • When seeking generic AI writing styles rather than specific personas.
  • When the primary need is for real-time persona generation without historical context.