Letting your AI fix things safely

Most things a scan finds are quick to fix in code or in your host's settings. Your coding assistant can often do it in a minute, if it gets good instructions.

Give it the fix prompt

Every thing to fix in Lockpick comes with a fix prompt. It says what was found, where, the evidence, what a good fix looks like and how to check it. Copy it and paste it into Claude, Cursor or whichever assistant you use.

If you've added Lockpick to your AI, it can read the finding itself and even check again when it's done.

Keep it on a short leash

• One finding at a time. Small changes are easy to review.
• Read what it changed before you publish. A good fix touches a few lines, not half the project.
• Say no to anything that disables a check instead of fixing the problem.
• Never paste secrets into the chat. If a finding is about a leaked key, replace the key yourself with the service that issued it.

Then check it worked

Publish the change, then use Re-read this record on a public-record finding, or Check again on your site for a new scan. A fix counts when the scan looks again and no longer sees the problem.