Shared site knowledge

Every read teaches Frankensurf something about a site: which tool got through, and which didn’t. That stays on your machine (route memory). Sharing it is opt in, both ways, and it’s only data: which tool reads a site reliably, never code and never anything signed in.

Start from the community’s

frankensurf share sync      # fetch community/routes.json from the repository
frankensurf share remove    # stop using it

sync downloads the community recipes, checks every one, and saves them beside your own recipes. A community recipe is a provisional seed: it orders the first tools tried on that site. It can’t turn on paid tools, use an identity, or override your own recipes and route memory, and a site a bundled recipe already covers keeps that recipe.

Share yours

frankensurf share export --out my-recipes.json

The export lists sites where one tool read reliably when the cheapest tool didn’t: at least 3 reads, at least 80% of them good. It includes only:

• the site’s origin and a short path pattern, cut to the first path segment, or the whole site when even that looks item-specific, so it says which sites need which tool, not which pages you read;
• public reads only, so nothing read with a profile, identity or your own browser;
• how many reads back each recipe.

Nothing is sent anywhere. Read the file, delete anything you’d rather keep to yourself, then open a pull request adding it to community/recipes/. Accepted recipes are merged into the file share sync downloads.

Note

Every file in community/ is checked by the same rules share sync applies: known recipe fields, a community_ id, single pages, and no authority settings such as identities, paid tools or provider lists.