An space for sharing your go-to agent skills so everyone can use them.
My list:
- Superpowers (planning & implementation)
- Caveman (for reducing token usage)
- Frontend Design (to help with styling)
- Remember (memory)
I have one i created myself called teachme, which uses a bit of thinking to explain topics to you and then it evaluates your understanding.
:)
Top comments (10)
My favorite agent skill is pattern recognition across multiple data streams.
I use AI to analyze sleep, recovery, workouts, weight trends, and daily habits together instead of looking at each metric in isolation. The most useful outcome isn't prediction, it's context. It helps identify what's likely water retention, training fatigue, poor sleep, or a real trend worth paying attention to.
That workflow eventually became the foundation of a project I'm building around GLP-1 weight-loss maintenance.
That's very cool!
Where do you get the data from? Custom tooling, Apple/Samsung/Etc Health?
Thanks! Right now the data comes primarily from Fitbit, plus manually tracked weight and health metrics. I'm experimenting with AI as a pattern-recognition layer that connects the dots across multiple data streams rather than looking at each metric in isolation.
Don't have anything that shares data with Fitbit, but if you ever manage to do with Apple Health, count me as a beta user 🎉
Thanks, Kevin! It's actually not tied to any specific platform or wearable. The idea is that people bring whatever data they're already tracking, weight, sleep, workouts, recovery, habits, notes, or other health metrics and use AI to connect the dots between them.
I'm exploring it primarily through the lens of GLP-1 weight-loss maintenance, where understanding patterns & context often matters more than any single data point.
Oh that's even better! Good luck on what you're doing.
If I can help u with somthing lmk
Thanks, Kevin, I appreciate that! I'm still very early and figuring things out as I go. It's been interesting translating a personal AI workflow into a method.
I'll definitely keep that offer in mind, and likewise, if I can ever be helpful from the GLP-1 side of things, feel free to reach out. Thanks again!
The teachme one is the most interesting to me, the eval step especially. Most "explain this to me" prompts just hand you an explanation and you nod along without knowing if it actually landed. Quizzing you after closes that loop. How does it decide you've understood, does it grade what you write back or just check you can restate the idea? And does it ever push back when you're confidently wrong, or take the restatement at face value?
Some comments may only be visible to logged-in visitors. Sign in to view all comments.