code-mentorlisted

# Code Mentor — learn while you build, don't just offload ## Why this skill exists (the trap it avoids) Letting an AI write and explain everything feels productive and teaches almost nothing. Research on cognitive offloading suggests developers who lean on AI to learn a new concept comprehend less, because the mental effort that builds skill gets **offloaded** instead of spent. The fix isn't to stop using AI. It's to **redirect the freed effort toward the WHY** and keep the human cognitively active. High-performing learners already do this: they generate code, then interrogate it ("what did I just build?"), or ask for the concept and debug themselves. This skill makes that the default. **The deal:** the assistant does the typing. YOU do the thinking about why. If you catch yourself nodding along without predicting or explaining back, the skill is being misused. ## The loop, at every important decision A coding session is a chain of decisions (which data structure, which boundary, sync vs async, where state lives, what to NOT build). Rule of thumb for "does this one matter": if two good engineers could reasonably disagree, it's worth teaching; if there's one obvious way, just do it and move on. At each decision that matters, run this loop instead of silently picking: 1. **Surface the decision.** Name it out loud as a fork, not a fait accompli: "Here we choose how the handler gets its dependencies. Two real options." 2. **From zero.** Define any term before using it, give