Skill Anatomy

What makes up a skill on disk — directory structure, required files, and optional assets

~/.hermes/skills/ <skill-name>/ SKILL.md required frontmatter + skill body instructions for the agent references/ optional read-only data the skill can reference templates/ optional file templates the skill can produce scripts/ optional executable scripts invoked by skill tools assets/ optional binary assets (images, data files) SKILL.md — frontmatter name display name on Skills page required description short summary required triggers phrases that invoke skill optional version tags dependencies semver string filter labels other skill names optional SKILL.md — body Instructions (Markdown) Natural-language instructions the agent follows when the skill is invoked. May reference templates/ and external tool calls. Skill Origins Built-in shipped with Hermes Agent or pre-installed always trusted Marketplace installed from Skills Hub reviewed by hub team varies by skill Agent-created generated by the agent during a session review before enabling User-authored written manually and placed in skills dir full trust (you wrote it) Optional directories — what they hold references/ docs, data files read-only at runtime templates/ Markdown / code scaffolds, snippets scripts/ shell / Python scripts elevated trust required assets/ images, binaries referenced by skill Security risk score computed from skill content • safe / low / medium / high shown in detail drawer before enabling Legend Required Optional (read-only data) Optional (executable) Security

SKILL.md (required)

  • • YAML frontmatter: name, description, triggers, tags, version, dependencies
  • • Markdown body: instructions the agent follows when invoked
  • • All other files in the skill directory supplement this file

Optional directories

  • • references/ — read-only context data the agent can cite
  • • templates/ — file scaffolds the skill can produce
  • • scripts/ — executable scripts (elevated trust required)
  • • assets/ — binary files referenced by the skill

Skill origins

  • • Built-in — pre-installed with Hermes Agent, always trusted
  • • Marketplace — installed from Skills Hub, hub-reviewed
  • • Agent-created — generated during a session, review before enabling
  • • User-authored — written manually, full trust

Hermes Switch UI • skill anatomy • SKILL.md manifest + optional directories + origin types