Skills#

Skills extend agent capabilities. They are declared in the YAML config and automatically deployed to each agent's .claude/skills/ directory at deploy time. Skills are the primary mechanism for inter-agent communication, identity, and other core capabilities.

Skills#

Skills use the standard Claude Code skill format -- a SKILL.md file with YAML frontmatter inside a named directory. Drop any skill into skills/ and reference it by name in your config.

Built-in Skills#

Skill	Description
`identity`	Agent name, team, and onchain ENS domain
`inter-agent`	Messaging, delegation, news feed for multi-agent coordination
`catalog`	REST-AP self-description visible to other agents
`wallet`	OWS multi-chain wallet addresses (skipped if no wallet)
`xmtp`	Encrypted messaging via XMTP network -- send and receive messages by ENS name (see XMTP Messaging)

Adding a Skill#

Create a directory in skills/ with a SKILL.md file:

skills/my-skill/
  SKILL.md

Add YAML frontmatter to SKILL.md:

---
name: my-skill
description: What this skill does. Claude uses this to decide when to invoke it.
allowed-tools: Bash
---

# My Skill

Instructions for the agent...

Reference it in your config:

defaults:
  skills:
    - identity
    - inter-agent
    - catalog
    - wallet
    - my-skill

Skills from defaults and per-agent lists are merged (deduped). You can download skills from Anthropic or the community and drop them into the skills/ directory.

Per-Agent Skills#

Add skills to specific agents in addition to the defaults:

agents:
  - name: researcher
    skills: [custom-research-skill]
  - name: coder
    skills: [code-review-skill]

How Skills Are Deployed#

At deploy time, each skill's SKILL.md is copied to the agent's working directory at .claude/skills/<name>/SKILL.md -- the standard Claude Code location. Template variables ({{VAR}}) in the skill content are substituted with deploy-time values like the agent's name, team, and domain.

Operational Skills#

These skills are available in the skills/ directory for manual use but are not deployed to agents automatically:

Skill	Description
`admin-control`	Manage the agent team from any Claude Code session -- remote commands, conversational messaging, and parallel work dispatch with polling. See CLI Reference -- Admin Control Skill
`local-agent`	Spawn and manage local Claude Code agents
`inter-agent-communication`	Shell scripts for agent-to-agent communication
`restap-client`	Generic REST-AP client for testing
`polling`	Async reply monitoring

To use these, reference them by name in your config's skills list.