Function validate_supernode 

pub fn validate_supernode(sn: &SupernodeConfig) -> Result<(), Vec<String>>

Validates a supernode definition’s structure, collecting all violations.

Enforced rules (spec §3):

• exactly one boundary node of type input, id input; one or more output boundaries (id = port name, output = the success port); one or more error boundaries (id = error-kind port name, error = the default the black-box rule targets);
• output/error ids must not be in RESERVED_OUTPUT_IDS/RESERVED_ERROR_IDS (collide with fixed instance port names like success, input, etc.);
• no two nodes share an id (duplicate ids would create ambiguous output ports);
• inner nodes must not be listener/client/supernode, must not use reserved ids, and must not contain /;
• exactly one edge leaves input; no edges into input or out of output/error boundaries;
• every edge endpoint references an existing node;
• no orphan inner nodes (unconnected output/error boundaries are fine — not every subgraph uses both exits);
• every inner edge leaves a port its source node’s type actually declares (out normalizes to success; boundary pseudo-nodes are exempt — they have no plugin type and therefore no PortSpec);
• every success/outcome port of every inner node is wired inside the definition, to another inner node or to a boundary. Definitions are compiled as part of each policy that instantiates them, so an unwired outcome port would otherwise surface as a confusing compile error on an unrelated policy rather than at save time on the definition itself. error ports stay exempt: the black-box rule in expand_policy wires every unhandled inner error port to the instance’s error exit.