# Save instructions you reuse

If you keep typing the same instructions, save them once. ChatGPT and Claude both let you keep instructions in a project, and some plans let you save a skill: instructions the AI uses only when the job matches. Saved instructions make results more consistent, but you still check them.

## What to do
Download the FAQ starter from Skill Labs. Put its instructions in a project in your AI app, or as a skill if your plan has skills. On a free plan, paste them at the start of a new chat. Test it twice: give it a normal FAQ, then one containing a made-up claim. If you saved it as a skill, test once more with a request that has nothing to do with FAQs. Then improve the instructions based on what happened.

## It worked if…
The instructions point out the claim they can’t back up. If you used a skill, it stayed out of the unrelated request.

## Your notes
- Which tool you used, and its version:
- What you started with:
- What you expected to happen:
- The steps you took:
- What actually happened:
- Where your screenshots are:
- What went wrong, and what you changed:
- What happened when you tried again:

Official guide: https://github.com/anthropics/skills

## Note down
- Where you saved the instructions:
- The FAQ you tested:
- The made-up claim it flagged:
- What you changed in the instructions:
