Agent kits by Rob Race
Give your AI agent a workflow, not another vague prompt.
Each kit gives Claude or Codex the skills, rules, and output formats for one specific job. You can inspect how it works, change it to fit you, and keep every file it creates.
What is an agent kit?
A practical package you install into the agent you already use. It turns a repeatable piece of work into a clear process instead of making you rebuild the prompt every time.
Building now · GitHub-first release
Engineering Evidence
See the engineering work your review will otherwise miss.Pull together work scattered across GitHub and Jira, preserve who did what, and build source-linked material for self reviews, promotion packets, and client updates.
For engineers, managers, and consultants · Creates a local dashboard, evidence packet, and review draft ↗Planned next · Free voice tool first
Personal Voice
Stop teaching every new AI conversation how you write.Turn approved writing and the edits you keep making into a voice profile and practical instructions that travel with you across Claude and Codex.
For technical professionals, founders, and consultants · Creates a voice profile, surface-specific rules, and agent instructions ↗How it works
- Use your agent.Bring Claude, Codex, or another capable model you already pay for.
- Install the workflow.Give it a tested process, clear boundaries, and useful output formats.
- Own the result.Read, edit, and keep the generated files instead of locking the work inside another app.