Rob Race

Agent kits by Rob Race

Give your agent durable context for the work you repeat.

Each kit turns a repeated piece of work into an inspectable workflow your agent can follow. Bring Claude or Codex, keep the files local, and own the context instead of rebuilding it in every conversation.

What is an agent kit?

A practical package of skills, rules, and readable working files for one specific job. The useful part is not a giant prompt. It is durable context your agent can reuse, you can inspect, and both of you can improve over time.

01

Early access · Available now

Personal Voice

A voice profile your agent can use—and you can inspect.

Build it from writing you approve, use it to polish drafts in Claude or Codex, then turn the edits you still make into proposed rules you control.

Local-first · Codex + Claude Code · No hosted writing corpus

How it works

  1. Bring the context.Choose the writing, examples, or working material that should teach the workflow.
  2. Use your agent.Run the workflow with Claude, Codex, or another capable model you already use.
  3. Own the result.Inspect, edit, and keep the generated context instead of locking the useful part inside another app.