What setup asks
The questions setup asks, the files it writes, and the one rule every session depends on.
Setup is the init skill. The Quickstart's guide ends with the line that starts it. By hand, run /agentlayer:init from a fresh Claude Code session in the agent's folder. Under Codex, create the folder from Claude Code first; see Codex.
Before it asks anything, setup checks the prerequisites and stops with the exact fix if one is missing. Then it installs the plugin's packages if they are not there yet.
The questions
A few questions about the agent, about you and your work, and about how it should run. Every question past the name can be skipped, and each answer becomes the default for the steps after it.
| About | Setup asks |
|---|---|
| The agent | Its name (default Oppi, plus an emoji, colors and an avatar if you rename it). Whether you start from a seed bundle. Its character. Its role: general, coding, research, planning or your own. |
| You | Your name and home timezone. Where your code lives. Links about you (blog, GitHub, talks). An avatar. How it should talk to you, and any hard preferences. Signal topics. |
| Your world | Your current projects, this year's goals, the company the agent works for, and documents to read. |
| How it runs | Whether knowledge/ and projects/ should live somewhere else. The model: Opus or Fable. Whether to keep a history of its memory. Whether you also run Codex from this folder. Which packs to switch on. |
Then it shows a summary and writes the files. The terms, in plain words:
- Name. Every file setup writes is phrased in it. See Naming your agent.
- Seed bundle. A zip from a teammate's agent that answers most of these questions for you. See Seed bundles.
- Character and role. How the agent talks and what it is for, saved as
SOUL.mdandIDENTITY.md. - Signal topics. The subjects the morning briefing searches for news on. Setup proposes a starter set.
- History. Saved versions of the agent's memory, so a change can be undone. See History.
- Packs. Groups of tools that switch on with a key, such as search or GitHub. See Connect it.
Your world
These shape what the agent knows about you from its first session.
Your projects, this year's goals, your company and the documents you hand over each become a file in the home; open the table for which file each one becomes.
LinkedIn blocks automated reads, so a profile link gives the agent nothing. Open your profile, choose More → Save to PDF, and drop the file in the inbox when setup asks. It's the fastest way to hand over your career history.
What it writes
Setup writes the operating manual, the identity files, the brain and work folders, and each host's settings; the table says what each one holds.
Every file and what it holds is on Agent home.
Setup never asks for a key in conversation, because a pasted secret lands in the transcript. You add keys in your editor afterwards; see where keys live.
If you moved knowledge/ or projects/ outside the folder, setup adds the permission entries the host needs to reach them without prompting.
Exit and relaunch
The plugin's registration only takes effect in a fresh session, so setup ends by asking you to exit and start again.
/exit
cd ~/Documents/Agents/Oppi && claudeWatch for a marketplace trust prompt on the first relaunch and accept it. If you miss it, the plugin does not load and no banner appears. Recover with /plugin install agentlayer --marketplace AgentLayer1/agentlayer-agent-marketplace.
The first session opens with a welcome. Your first ten minutes picks up there.
The one convention
Always launch from the agent's folder (its home), then reach your code from inside the session.
The plugin is switched on by the home's settings, so a host loads it only for sessions started there. That one fact gives you the agent's identity, the skills, the tools and session capture. It is also what keeps several agents on one machine apart with no configuration.
Reaching your code. Add the repos you work in to permissions.additionalDirectories in the home's .claude/settings.local.json. Both hosts honour it: Claude Code grants file access there, and Codex lists the same paths as workspace roots.
{
"permissions": {
"additionalDirectories": ["/Users/you/Developer/acme", "/Users/you/Developer/side-project"]
}
}Set AGENT_CODE_PATH to your primary repo and its git activity shows up in every session's context.
The session can wander, the anchor cannot. Once running, cd anywhere. The home is resolved from the launch directory, which never moves, so capture lands in the right brain even when the shell has roamed three repos deep.
What you give up
A session started outside a home is not captured, and says nothing about it. No plugin loads, so no hooks run. Do not put AGENT_HOME in a user-level settings file to work around this: it is machine-wide and hands every session to one brain. Per-home config belongs in that home's settings.local.json.