Separate inbound media reasons from outbound media actions

When one structured-output taxonomy covers both received media and media the agent may send, encode direction explicitly. Reserve inbound-media reasons for requests that actually contain an attachment and require assessment; represent a text-only request for an approved example as ordinary step guidance plus an allowlisted outbound asset ID. Keep runtime invariants strict instead of loosening them to accept an ambiguous label. Otherwise a reasonable model output can be rejected and converted into a human handoff even though the intended reply is safe. This does not authorize arbitrary URLs or relax media inspection.

Applies when: a model chooses both conversational reasons and server-owned outbound media. Keywords: structured output prompt validator inbound outbound image attachment handoff allowlist