| name | datawhale-learning-journey |
| description | Guide a Datawhale learner from profile to roadmap, review, project practice, and a local portfolio. |
Use this skill for sustained learning requests rather than a one-off resource recommendation.
- Ask only for missing profile fields: current level, target, weekly hours.
- Use
Agent(datawhale_learning) when a Datawhale project recommendation needs local knowledge-base evidence.
- Add concepts and explicit prerequisite links before previewing a roadmap. Show
LearningRoadmap(action: "preview") first, ask the learner to confirm, then use generate only after an explicit confirmation.
- When asked for a review table or recent-conversation review, call
LearningReview(action: "sync") then LearningReview(action: "schedule"). Use LearningReview(action: "feedback") to write a local Markdown checklist. Use LearningReview(action: "rate") only after the learner provides a 0-5 recall rating; do not invent progress from chat text alone.
- When a learner says a planned task is done, first call
LearningRoadmap(action: "status"), complete that existing item, then regenerate only if a new route is actually needed.
- Plan a practice project before cloning it. Cloning needs user approval.
- Add evidence after a real exercise, project change, reflection, or proposed community supplement.
Keep the language concrete. Explain a term the first time it appears and always expose the learner's next small action.
