id

Skills

Extensible agent capabilities through the skills system

Skills#

Skills are packages of instructions and executable scripts that extend what agents can do. They follow the Agent Skills Open Standard and are compatible with Claude and other AI platforms.

What Are Skills?#

A skill is a directory containing a SKILL.md file with instructions, plus optional scripts, templates, and data files. When an agent loads a skill, it gains the knowledge and tools needed to perform specialized tasks such as communicating with other agents, managing a cluster, or interacting with external services.

Available Skills#

inter-agent-communication#

Enables agents to discover and communicate with other agents in the cluster.

Use cases:

  • Delegate tasks to specialized agents
  • Coordinate multi-agent workflows
  • Get second opinions or validation
  • Scale work across multiple agents

Included files:

  • SKILL.md -- Instructions and examples
  • list-agents.sh -- List all available agents
  • talk-to-agent.sh -- Send a message to an agent
  • wait-for-response.sh -- Wait for an agent's response
  • broadcast-to-agents.sh -- Send a message to multiple agents
  • pay-agent.sh -- Payment system integration

admin-control#

Enables an agent to act as an admin for remote management of the ID Agents manager.

Use cases:

  • Remote control of the agent cluster
  • Execute CLI commands via API
  • Monitor and manage agents programmatically
  • Build automation and orchestration workflows

Included files:

  • SKILL.md -- Instructions and examples
  • talk-to-manager.sh -- Send a message to the manager with a reply endpoint
  • remote-command.sh -- Execute CLI commands with an API key
  • start-listener.js -- Start a temporary HTTP listener for replies

local-agent#

Spawn and manage local Claude Code agents using your existing authentication.

Use cases:

  • Use your Claude Code login instead of API keys
  • Debug and develop locally without containers
  • Run agents with full filesystem access
  • Participate in team workflows alongside containerized agents

Included files:

  • SKILL.md -- Instructions and examples
  • spawn-local.sh -- Spawn a local agent from the command line

restap-client#

A generic REST-AP client for testing and interacting with agents via the protocol.

Use cases:

  • Test REST-AP endpoints from the command line
  • Debug agent communication issues
  • Quick integration testing
  • External client simulation

Included files:

  • SKILL.md -- Instructions and examples
  • talk.sh -- Send a message to an agent
  • news.sh -- Get news and replies from an agent
  • discover.sh -- Fetch an agent's discovery document
  • listen.sh -- Start a temporary listener for replies

Using Skills#

As an Agent#

Agents access skills by reading the documentation and executing the scripts:

# Read the skill instructions
cat ./skills/inter-agent-communication/SKILL.md

# List available agents
cd ./skills/inter-agent-communication
./list-agents.sh

# Talk to another agent
./talk-to-agent.sh "coding-agent" "Create a button component"

In YAML Configuration#

Skills can be assigned to agents through the deployment config. For Claude Code agents, use plugins:

defaults:
  plugins:
    - name: id-rest-ap
      path: plugins/id-rest-ap

agents:
  - name: coder
    plugins:
      - name: id-rest-ap
        path: plugins/id-rest-ap
      - name: git-tools
        path: plugins/git-tools

Skills can also be provided as .md files:

agents:
  - name: my-agent
    skills:
      - name: inter-agent
      - name: custom-skill
        path: /path/to/custom.md

Creating a New Skill#

  1. Create a directory under skills/:
skills/your-skill-name/
  1. Add a SKILL.md file following this structure:
# Skill Name

## Overview
What the skill does.

## Available Operations
What actions are possible.

## Usage Examples
Concrete examples with code.

## When to Use
Guidance on when to apply the skill.

## Best Practices
Tips for effective use.

## Important Notes
Warnings, limitations, and considerations.
  1. Optionally add executable scripts, templates, and data files.

Skill Directory Structure#

skills/
├── your-skill-name/
│   ├── SKILL.md           # Main instructions (required)
│   ├── script.sh          # Executable scripts (optional)
│   ├── template.txt       # Templates (optional)
│   └── data.json          # Data files (optional)

Future Skills#

Planned additions to the skills library:

  • file-sharing -- Share files between agents via workspace
  • task-coordination -- Coordinate complex multi-agent tasks
  • result-aggregation -- Combine results from multiple agents
  • agent-monitoring -- Monitor agent health and progress
  • resource-management -- Manage CPU, memory, and cost across agents