When you hit a bug, your first instinct is to dive into the code and start fixing what you think is wrong. This skill enforces the opposite: check what the system actually produced first. Look at the directories, read the logs, run the failing command manually. The pattern comes from real debugging sessions where the issue wasn't what it seemed. A token limit error hiding in output, looking in the wrong cache path and assuming code didn't run, that kind of thing. It's basically "measure twice, cut once" for debugging. Check the filesystem and logs before you touch a single line of code, because outputs don't lie even when your assumptions do.
npx skills add https://github.com/parcadei/continuous-claude-v3 --skill observe-before-editing