Problem
AI learning resources move quickly, but trend lists alone do not help a Windows-based learner inspect Skills, install them safely, or keep a durable learning path.
My role
I designed the content model, bilingual learning path, local-installation safety guidance, Skill browsing flow, and deployed web experience.
Constraints
Recommendations must stay useful as tools change, avoid unsafe one-line installation habits, and work for learners who use Windows terminals.
Key decisions
The experience connects each trend to prerequisites, inspection steps, guided practice, batch install/download paths, and Agent learning notes instead of treating popularity as mastery.
Architecture
A static-first Astro surface separates editorial content, ranked Skill data, safety notes, and reusable learning modules while remaining inexpensive to host.
Engineering evidence
The deployed route, real learning modules, Skill browsing data, installation checks, bilingual content, and Agent book entry provide inspectable proof.
Result
HeatStack acts as a durable learning asset rather than a disposable trend post, with a route from discovery to practice and proof.
Limits & next steps
Trend freshness depends on the upstream collection process, and individual Skill quality still requires human judgment before recommendation.