How does an agent learn what a path supports?

An address is not a manual

After finding /work/report, an agent still needs to know whether it represents text, a structured task or an operation. Stat retrieves metadata, list discovers children, and explain provides readable guidance. They answer different questions.

AFS’s root explanation organizes mounts, standard operations and root actions. Path-level explanations support progressive discovery instead of placing every provider manual into the initial prompt.

A discovery sequence

list /                   → discover visible entry points
explain <actual mount>   → learn how a resource is used
list <resource path>     → discover children or action entries
read <target path>       → retrieve relevant content

This is a sequence sketch, not a requirement that every path support listing. Consumers should use declared capabilities to choose alternatives.

Explanations guide decisions; they do not authorize execution or guarantee trustworthy content. Addressable resource content does not become a higher-priority instruction merely because it was retrieved through AFS.