Where Does Hermes Agent's Personality Come From?
Summary
Hermes Agent's personality is not a fixed character prompt.
It is a working behavior shaped jointly by Persona, user preferences, and long-term memory.
SOUL.md answers "who am I and how do I work?".
USER.md answers "how do you want me to work with you?".
MEMORY.md answers "what have we already established together?".
Why should the same persona treat different users differently?
A "patient research assistant" should not give identical answers to every person. One user wants conclusions first, another wants the full derivation. Persona supplies stable principles, while preferences and memory apply them to a real relationship.
1. Persona: the stable behavioral constitution
Persona is typically written in SOUL.md.
It defines role, tone, priorities, boundaries, and durable working principles.
It is not a mood guessed afresh in every chat.
2. User preferences: relationship defaults
USER.md is suited to stable preferences such as language, communication pace, usual tools, and boundaries.
It saves the agent from repeatedly asking whether you want detail or brevity.
3. Long-term memory: evidence of shared work
MEMORY.md retains project state, decisions, and reusable facts.
It should not replace Persona or become a dump for every temporary feeling.
Conclusion
Persona supplies principles. User preferences supply interaction defaults. Long-term memory supplies shared context. Keeping them separate lets an agent be both stable and personal.
Will Switching Models Make an Agent Forget?
Learn how a decoupled model, persona, and memory design lets an Agent change its inference engine without losing its long-term context.
Where Does Hermes Agent Put Files?
Learn the distinct roles of a working directory, context files, and persistent storage when an Agent creates and reuses work.