"Explain when to keep state in the component vs. in a global store."
The skill names the fork ("state ownership"), defines terms from scratch, asks for your prediction before going further, then explains the real mechanism (lifecycle, coupling, invalidation). It illustrates with the minimum viable code and asks you a question back so you have to reformulate. The session is saved as a reusable lesson.
What it does
At each real decision (data structure, boundary, sync/async, where state lives), a loop: name the choice as a fork, define terms from zero, make you PREDICT before revealing, explain the real mechanism (not the surface), show minimal code, then hand the thinking back (explain this line to me). It calibrates to your level and fades the scaffolding once you master it. Output: a markdown file that reads like a lesson (problem, options, the call and why, example, self-check question, recap + review date).
When to use it
When you want to understand and progress, not just receive code. Stack-agnostic. Not for when you want the code fast (normal session).
The stance, research-grounded: letting the AI write and explain everything ships faster and teaches far less (cognitive offloading, ~17% lower comprehension, Anthropic study). The fix isn't to stop using AI, it's to redirect the freed effort toward the WHY and keep the human active (predict-before-reveal, teach-back), the way high performers do. The AI types, you think.
Install this skill
$ /plugin marketplace add ohugonnot/claude-skills
$ /plugin install code-mentor@web-developpeur-skills
Frequently asked questions
When should I use this skill instead of a regular session?
When you want to understand a mechanism, not just move the code forward. If you're in a rush and need the patch as fast as possible, stay in a normal session. This skill deliberately slows the flow so you work through the why.
Why does it ask me to predict before explaining?
Letting the AI write and explain everything ships faster but teaches much less (cognitive offloading, roughly 17% less comprehension according to an Anthropic study). Predicting first redirects that freed-up effort toward reasoning rather than passive consumption.
What does the saved trace actually look like?
A structured file in lesson format: named forks, definitions, minimal code, your answers to the follow-up questions. Reusable later as a review note.
Feedback & discussion · 0
No feedback yet. Tried it? Tell us what you think.