Post

our guides go out as resources today, so this doesn't bind us yet. if we served them as skills it would. we edit them often, and every edit changes a digest and re-prompts the user. first spec i've read that argues for batching your edits.

Curious how this compares to MCP prompts. Those can orchestrate as well and use roles. Is the intent here that skills can be discoverable and frontmatter loaded into context automatically after initialization? Looking through the docs like github.com/modelcontext... I don't see mention of prompts.

Thank you for the consistent really high-signal feedback!

🙏🏻