this is pure f*cking treasure
these 10 agent skills have 8.01M combined downloads and form a complete working stack
discovery. pressure-testing. design. browser work. prototyping. debugging. vision. orchestration. tool building
01 find-skills
▸ github.com/vercel-labs/sk…
02 grill-me
▸ github.com/mattpocock/ski…
03 frontend-design
▸ github.com/anthropics/ski…
04 agent-browser
▸ github.com/vercel-labs/ag…
05 prototype
▸ github.com/mattpocock/ski…
06 diagnosing-bugs
▸ github.com/mattpocock/ski…
07 skill-creator
▸ github.com/anthropics/ski…
08 image-to-code
▸ github.com/leonxlnx/taste…
09 subagent-driven-development
▸ github.com/obra/superpowe…
10 mcp-builder
▸ github.com/anthropics/ski…
the stack covers the full loop:
find the capability shape the idea build it inspect it verify it and package the workflow for the next run
save this, then select your gpt-6 astra use case
GPT-6 Astra Use Cases: 15+ Business & Developer Applications You Can Steal
For years, enterprise AI followed a relatively simple model: an employee entered a question and an AI system returned an answer. That model remains useful, but it leaves most of the actual work to the...
GPT-6 Astra Use Cases: 15+ Business & Developer Applications You Can Steal
For years, enterprise AI followed a relatively simple model: an employee entered a question and an AI system returned an answer. That model remains useful, but it leaves most of the actual work to the...
[댓글 @diamai_]
I would start with diagnosing-bugs before handing an agent the rest of the stack.
[댓글 @DGlushakov41949]
Strong shortlist. I would add an eval layer before chaining these skills, with clear pass criteria and trace logs. Otherwise, orchestration can make a small failure much harder to locate.
[댓글 @notp3rk]
grill-me sitting right next to the anthropic official skills is a fun contrast in what actually gets used
