The problem
Documentation was spread across working notes, project folders, a wiki and many separate Claude sessions. In a real emergency I needed to know where everything was without guessing under stress.
What I did
- Built a single index of every system showing its documentation, recovery source and what is still missing, with each one marked as having a step-by-step emergency runbook, reference docs only, or nothing yet.
- Added a table that maps each emergency scenario to the right document.
- Made the index part of every Claude session: read a system’s row before working on it, update it and the change log after. Links only, never secrets.
- Pulled four long-running recovery and security planning threads into one entry point with a single list of open items.
- Built a 72-page printable offline pack of the 11 recovery runbooks and the index, with no secrets in it, for the emergency binder.
- Wrote knowledge-base articles in a consistent Markdown house style, more than 70 by September, and published them to the internal wiki.
Outcome
One place shows where every system’s documentation lives and which systems still need a runbook. The recovery runbooks also exist as a printable copy that doesn’t depend on the systems it describes.