AI Guides › Skills & Agents

Auditing Your Installed Skills Once A Quarter — What To Keep, What To Delete

By Nigel Guy · 2 min read

Skills accumulate quietly. You install one for a task, it works, you forget about it, and six months later you've got a library of tools you don't remember the details of, some of which are quietly out of date with the platforms they connect to. Nobody notices until one of them breaks in a way that costs something.

The rule: review your installed skills on a fixed cadence, not when something breaks — a quarterly audit catches rot before it costs you anything.

The mechanism

  1. List everything installed, not just what you remember using. The forgotten ones are exactly the risk this audit exists to find.
  2. For each one, ask: have I used this in the last month? If not, either there's a good reason (seasonal, occasional) or it's a candidate for removal.
  3. Check whether the tools it depends on have changed. An API update, a platform redesign, a renamed field — any of these can silently break a skill that used to work fine.
  4. Re-read the guardrails section, if it has one. Does it still match how you're actually using it, or has the use case quietly expanded beyond what the guardrails were written for?
  5. Delete, don't just ignore, anything that fails this check. An unused skill isn't neutral — it's an untested, possibly-rotted piece of automation sitting in your library.

What to skip

Skip auditing weekly — that's more overhead than the risk justifies for most personal use. And skip keeping a skill "just in case" once it's failed the audit twice running; if you haven't needed it in six months, rebuilding it later is cheaper than maintaining it indefinitely.

Guardrails

All 751 AI guides · JulieMango plans from £17/mo