Getting Started

This profile, and how to extend it

A Hermes agent is three things in one directory:

{{agent_package}}/
├── SOUL.md        identity: who I am, what I'm for, how I speak
├── config.yaml    model, max_turns, platforms, enabled skills
└── skills/
    └── getting-started/
        └── SKILL.md      ← this file

SOUL.md is loaded every turn — it's short on purpose. config.yaml is operational, not behavioural: changing model.default changes which model answers, never what the agent is.

What the agent sees

At startup only the frontmatter of each skill is loaded — the name and description above. The body you're reading arrives only when this skill is actually opened.

That split is the whole design. Fifty skills cost a few hundred prompt tokens until one becomes relevant, then the full instructions show up exactly when they're needed. It's also why the description matters more than it looks: it's the only thing the agent has when deciding whether to read further.

Adding one

skills/
├── getting-started/
│   └── SKILL.md
└── your-new-skill/
    └── SKILL.md          ← add this; nothing else to wire

The frontmatter is two keys:

---
name: incident-triage
description: Triage a production incident — severity rubric, who to page, and what to capture in the timeline. Read when an alert fires or someone reports an outage.
---

Write the description for a reader deciding whether to open the file: name the situation, not the topic. "Read when an alert fires" beats "about incidents."

skills.enabled in config.yaml is a force-load list, not a registry — leave it empty and every skill in this directory is still discoverable. Add a slug there only when the skill must be in context from the first token.

Writing the body

The body is instructions for a capable colleague who hasn't done this particular task before: concrete steps, real examples, the failure modes worth naming. Skip the background theory — the model already has it.