Turn a manual runbook into a replayable agent skill

Runbooks that actually get followed

Record the incident or deploy runbook as you perform it. SkillForge captures each step into a SKILL.md that an agent can execute or a teammate can follow exactly.

A worked example

Take a "rotate a leaked API key" runbook that lives half in a wiki and half in someone's head. Instead of a stale doc:

1. Open skillforge.expert/app and click Record the next time you do it.
2. Walk the real steps: open the secrets manager, revoke the old key, generate a new one, update the config/deploy, and verify the health check.
3. Stop. SkillForge captures each click and input into a SKILL.md — a precise, ordered spec.
4. Now an on-call agent can execute it (with your review) or a new hire can follow it exactly, no tribal knowledge required.

Why recorded runbooks beat wiki docs

Written runbooks drift from reality and skip the "obvious" clicks. A recorded SKILL.md captures exactly what was done, in order, including the small steps people forget to document.

Related guides

See also QA test automation, SkillForge as an RPA alternative, and browser-use skills.

FAQ

Is this safe for production steps?
SkillForge produces a spec you review before any agent runs it — you stay in control of execution.

Can a human follow the SKILL.md instead of an agent?
Yes — it is human-readable Markdown, so it doubles as a precise runbook for new team members.

Does it capture terminal / CLI steps?
SkillForge captures on-screen actions from your recording, including what you type; review the generated spec and edit any sensitive values before sharing.