Post-v1: progressive discovery for unenumerable skill catalogs

[olaservo]

[aditya-scio]

@sambhav @SamMorrowDrums - started a thread here https://discord.com/channels/1358869848138059966/1464745826629976084/1539403520128847903

[SamMorrowDrums]

Some libraries of skills are fundamentally un-enumerable without progressive discovery of some form.

(Whether we would or not) GitHub could make available all skills in all repos, the same is it already does for files (including all skills.md files in accessible repos).

Whether it's resource templates, or some kind of graph based discovery, it's fair to say I expect this problem to be solved across the entire protocol in time.

GitHub has a big configurable MCP that can be small if you want a subset, Cloudflare went with many servers first, and then did their whole server side code mode thing.

For me these are all shades of the same problem, and I don't think skills needs to solve it alone, but should eventually subsume any solution that does emerge.

[olaservo]

Retitled as a post-v1 proposal. v1 dropped resource-template index entries (decision log, 2026-07-16), so this is now the broader question: how do hosts progressively discover skills from servers whose catalogs are unenumerable — templates + completions, server-side search, or a skills/list filter (#99). Per CONTRIBUTING.md, the vehicle is a decisions.md entry with Status: Proposed.