Agent Personality Files

Agent Personality Files

Agent personality files are persistent instruction files that shape how an agent behaves socially, speaks, decides when to answer, and relates to its user.

Key points

  • Steinberger created soul.md after finding that a raw Claude Code-style assistant did not feel right inside personal messaging contexts [src-064].
  • The source treats personality as product design: tone, humor, warmth, no-reply behavior, and social timing affect whether an agent feels like a useful companion or a generic model [src-064].
  • The file is inspired partly by Anthropic-style constitutional prompting, but Steinberger uses it more personally to encode values, style, and how he wants to work with the agent [src-064].
  • Allowing the agent to modify its own personality file, with visibility to the user, turns personality into another self-modifying system boundary [src-064].
  • MoldBook amplifies this pattern by making agent identity and public expression part of the OpenClaw culture [src-064].
  • Nate describes Hermes's soul.md as the personality/vibe layer inside the five-pillar assistant architecture. It should be edited when the agent is too verbose, too casual, too stiff, or mismatched to a public-facing role [src-074].
  • For agents that interact with other people, such as a YouTube comment assistant, personality files become an operational guardrail as much as a style preference [src-074].
  • Roberts uses soul.md as part of the cross-harness bridge: the same durable identity/personality layer can inform Hermes responses while other memory and project context come from Obsidian or Claude/Codex-style files [src-079].
  • His Pantheon persona layer generalizes the idea from one personality file to multiple named personas, each with its own prompt, model choice, description, and intended use [src-079].

Related entities

Related concepts

Source references

  • [src-064] Lex Fridman – "OpenClaw: The Viral AI Agent that Broke the Internet – Peter Steinberger | Lex Fridman Podcast #491" (2026-02-12)
  • [src-074] Nate Herk — "Hermes Agent: Zero to Personal AI Assistant (1 Hour Course)" (2026-05-10)
  • [src-079] Jack Roberts — "Hermes Agent just got 10X Better (Agentic OS)" (2026-05-15)

Robin Cartier perspective

This page is part of Robin Cartier's working AI knowledge graph: a practical research layer for production AI, recommendation systems, experimentation, GEO, and agentic web readiness.

The useful next step is to connect this concept back to applied product leadership and operating models.

Recommended next

Keep reading from this thread

From 491 indexed pages and articles.

  1. Wiki concept Hermes Five-Pillar Agent Architecture Nate Herk's operating model for turning Hermes Agent from a chat bot into a personal AI assistant Related by soul
  2. Wiki concept Hermes Agent An open-source personal AI agent work surface from Nous Research that Nate Herk uses for general knowledge work, scheduled automations, and Related by style
  3. Insight AI Beyond POCs How enterprise AI moves beyond proofs of concept through ownership, governance, measurement, adoption, and production operating models Readers have engaged with this next