Agents / Agents

Onboarding Author

Agent

Builds onboarding guides that get a newcomer productive fast — role-based first weeks, setup that verifiably works, the unwritten rules written down — tested against reality, not memory.

Best for

  • Teams where onboarding is 'shadow someone and ask questions'
  • Shrinking time-to-first-contribution from weeks to days
  • Capturing what only the senior person knows before they leave

What you give it

  • Access to the project, tools and existing docs; who the newcomer is
  • Answers to the questions only veterans can answer — it will ask precisely

What you get back

  • A day-one guide where every setup step was actually verified
  • A role-based first-week path: what to read, do and ship, in order
  • The tribal knowledge captured: how decisions happen, who owns what, the gotchas

How it works

  1. Tests existing onboarding material against reality first — broken steps are found by running them.
  2. Designs backwards from 'productive': the first real contribution defines the path to it.
  3. Interviews the veterans for the unwritten layer: norms, ownership, history, traps.
  4. Sequences by need-to-know, not everything-at-once — day one is not the place for the architecture lecture.
  5. Builds the feedback loop in, so each newcomer leaves the guide better.

Example

You: Our next engineer starts Monday. Onboarding is currently a wiki page from two years ago.

Result: The wiki page tested: 6 of 14 steps broken — fixed. A first-week path ending in a real shipped change on day four; a 'how things really work' section (review culture, deploy rhythm, who to ask what) built from interviewing you; and a day-five feedback loop so the guide improves with every hire.

Limits — please read

  • Verification covers what it can execute; tool access only humans hold gets checklist treatment.
  • Culture is captured as described plus observed; it cannot interview people you do not make available.
  • Guides rot like docs rot — the feedback loop slows it, maintenance remains real.