helping developers get more out of @sanity_io \n (he/him) \n ask me about real-time text-fields fields in front of delightful JSON \n advisor @heavybit
so, "end of day" according to @FedEx is 4pm and their delivery status page is not very good.
(a bit 😡 after having stayed at home for 2 days for a package that requires signature, doesn't get delivered, and having no means of knowing when they actually show up or access to any
this is the correct take - hence, we built trigger evals right into our "skills studio" at Sanity and require them to pass before you can distribute your skill internally.
this gives folks a feedback loop that helps them tune their skills descriptions etc.
Most agent skill evaluations focus on how a skill performs once it's loaded.
You also need to test whether your skill gets called in the first place. Most evals assume the skill is already active. However, coding agents only inspect the short name and description to decide
soon we'll be able to cross-post @sanity_io engineering posts on X, right from our studio!
→ Tables and mermaid charts (thanks to Pele by @kepano!) are automatically rendered to pngs since X's article composer doesn't support these formats¹
→ Header image is automatically
MCP Events are pretty cool. It's still a drafted spec, but ChatGPT has it already 👏
However, the holy grail of streaming support might not be coming to clients anytime soon (which is reasonable).
Still, my colleague @ai_ftw_ made an mcp events extension for @pidotdev to see
