Guide · Agent skill

How to use the Agent Starters agent skill (agent-starters)

Updated 2026-11-02 · written with AI from the project's own documentation and checked against our catalogue data

The agent-starters skill teaches an AI agent how to create, refresh, inspect, promote and deprecate a named starting point for an agent seat. A starter combines the seat's role, startup context, an optional native session source and provenance notes. This guide is for people who have never used agent skills and want to install this one, trigger it, and understand what it can and cannot do.

Agent Startersmvschwarz/openrig · Apache-2.0 Download the skill

Quick start

1Confirm your requirements. Check that you have the OpenRig daemon and the rig CLI, and choose a registry directory for starter files.
2Download and unzip the skill. Download the ZIP from this page and unzip it so the folder contains SKILL.md.
3Place the skill folder. Move it to ~/.claude/skills/ for personal use or to a project's .claude/skills/ folder, or upload it under Settings → Capabilities → Skills in the Claude apps.
4Write a starter entry. Create a YAML file with starter_id, role and context, and save it in your registry directory.
5Ask for the task in plain language. Request creating, inspecting, promoting or deprecating a starter by name, using the skill's terms.
6Use the starter in a member. Reference it in a rig member with starter_ref and name, then check the result with inspection.

What an agent skill is and what this one makes the agent do

An agent skill is a folder with a SKILL.md file that gives an AI agent instructions for a specific kind of task. When a request matches the skill's description, the agent follows its guidance.

This skill covers the full life of a starter. It can help the agent capture a useful seat pattern, name it with a stable id and owner, make its runtime, context inputs and provenance visible, and mark it promoted or deprecated. It defines a six-state lifecycle: captured, named, inspectable, used, promoted and deprecated.

Creating a new starter from a known-good seat context
Refreshing a seat with rig expand and starter_ref
Authoring a registry entry
Inspecting a starter's provenance, freshness or recommended status
Promoting or deprecating starters in the registry

Install the agent-starters skill

Download the ZIP from the skill's page in this directory and unzip it. The skill is a folder that must contain SKILL.md. Place it in one of these locations:

For a personal install, put the folder in ~/.claude/skills/. To limit it to one project, put it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.

Personal: ~/.claude/skills/
Project only: the project's .claude/skills/ folder
Claude apps: Settings → Capabilities → Skills (upload)

How to trigger the skill with a natural request

You do not need a special command. Describe the task in plain language and the agent should match it to this skill's description. The skill is written for creating, refreshing, packaging, inspecting, promoting or deprecating a starter, so use those words.

Name the starter and the seat when you can. Mention whether you want to look at an existing starter or make a new one, since the skill handles those differently.

Create a starter from the reviewer seat that is working well now.
Inspect the reviewer-v1 starter and tell me where its context came from and whether it is still current.
Deprecate the old reviewer starter because its doctrine is out of date.
Refresh this seat from its starter using rig expand.

Write a starter registry entry

A starter is saved as a YAML file in a registry directory. The example below is saved as reviewer-v1.yaml. The resolver uses an explicit root if you set one, otherwise OPENRIG_AGENT_STARTER_ROOT, otherwise the home registry at ~/.openrig/agent-starters.

The current resolver checks the entry shape and the credential rule, then delivers the YAML itself as one required guidance_merge startup file on a fresh start. It does not interpret arbitrary context references or load a native conversation from this example.

starter_id: reviewer-v1
role: Review the assigned change against its stated outcome.
context: Read the current task and the source needed to judge it.

Use a starter in a rig member

A rig member refers to a starter by name with starter_ref. This is the usage shown in the skill for seats created with rig expand.

A RigSpec member that combines starter_ref with session_source.mode: fork is currently rejected.
If you need native conversation continuity, the skill says to use a separate supported session-source path.

members:
  - id: reviewer
    starter_ref:
      name: reviewer-v1

Safety, requirements and limits

The skill does not run scripts and does not use the network. Its own files are only instructions. The commands and resolver it refers to belong to the OpenRig daemon, so the skill cannot do anything useful without that system installed.

The skill refuses to copy provider API keys or auth material into a starter. It refers to session sources and context instead. It also refuses to compact, summarize or shrink a seat just to make a smaller starter, since compaction is lossy.

Requires the OpenRig daemon and the rig CLI, including rig expand and starter_ref support.
Requires a registry directory for starter YAML files.
Runs scripts: no. Uses network: no.
Starters are not VM images and do not capture machine state.
Unsupported runtimes and stale or invalid session sources are refused with an error.
If provenance is not inspectable, the skill refuses promotion.

What people use it for

Reusing a known-good reviewer seat configuration across rigs
Checking whether a shared starter's source and context are still traceable
Marking an outdated starter as deprecated so teammates stop using it
Refreshing a seat from a saved starter with rig expand

Frequently asked questions

Is the agent-starters skill free and what license does it use?

The source lists the license as Apache-2.0. Check the project's repository for the full license text before you reuse or redistribute it.

What do I need to run the agent-starters skill?

You need the OpenRig daemon and the rig CLI with rig expand and starter_ref support, plus a registry directory for starter YAML files. The skill alone cannot run those commands.

Where do I install an agent skill like this?

Put the unzipped folder in ~/.claude/skills/ for personal use or in a project's .claude/skills/ folder. You can also upload it in the Claude apps under Settings → Capabilities → Skills.

Is an agent starter the same as a VM image?

No. The skill states that starters are not VM-style images and do not capture machine state. A starter is a managed starting point built from role, startup context, an optional session source and provenance.

Does the skill copy my API keys into a starter?

No. The skill refuses to copy provider API keys or auth material into a starter and refers to session sources and context instead.

Why does my starter get rejected when I combine it with session_source fork?

The skill says a RigSpec member combining starter_ref with session_source.mode: fork is currently rejected. Use a separate supported session-source path when you need native continuity.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Use this skill if you run OpenRig rigs and want reusable, traceable agent seat configurations with clear provenance and staleness reporting. Pick something else if you need deterministic VM-style state capture, or if you do not use the OpenRig daemon, since the skill cannot run without it.

Download the skill