Skip to main content

Create Your Own Personality

Drop a directory into ~/.ethos/personalities/<id>/ with three files. Ethos picks it up automatically — no restart needed.

Step 1 — Create the directory​

mkdir -p ~/.ethos/personalities/strategist

Step 2 — Write the identity​

ETHOS.md is read by the agent as part of its system prompt. Write it in first person. Be specific about communication style, what the agent prioritises, and what it avoids.

~/.ethos/personalities/strategist/ETHOS.md
I am a strategic advisor focused on long-horizon planning and prioritisation.

I help identify what matters most, what to defer, and what to drop entirely.
I think in terms of leverage: which actions compound over time, which are
one-time costs.

I ask clarifying questions before giving advice. I'm direct about tradeoffs.
I don't pretend decisions are easier than they are.

Step 3 — Configure the personality​

config.yaml uses simple key: value format — no nested YAML.

~/.ethos/personalities/strategist/config.yaml
name: Strategist
description: Long-horizon planning and prioritisation
model: claude-opus-4-7
memoryScope: global

Fields:

  • name — display name (shown in /personality list)
  • description — one-line description
  • model — LLM model to use for this personality
  • memoryScope — global or per-personality

Step 4 — Define the toolset​

List the tools this personality is allowed to use.

~/.ethos/personalities/strategist/toolset.yaml
tools:
  - web_search
  - read_file
  - memory

Available tools depend on your Ethos version. Run /help in chat to see what's installed.

Step 5 — Switch to it​

/personality strategist

:::tip Hot-reload Ethos watches config.yaml modification times. Edit any personality file and the changes are live on the next message — no restart required. :::

Tips​

Keep ETHOS.md in first person. The agent reads it as a description of itself. Third-person descriptions ("This agent is...") work but feel less coherent.

Keep toolsets minimal. Only include tools the personality actually needs. A reviewer with no write access is safer than one that can accidentally modify files.

Choose memoryScope deliberately.

  • global — use when this personality should share context with your other global-scope personalities (e.g., project notes, ongoing tasks)
  • per-personality — use when this personality's context should be completely isolated (e.g., a persona for reviewing sensitive documents)

Model routing per personality. You can run different models for different personalities — fast/cheap for coach, powerful for engineer:

# engineer/config.yaml
model: claude-opus-4-7

# coach/config.yaml  
model: claude-haiku-4-5-20251001