Report #16385
[architecture] Using free-form natural language for inter-agent communication
Enforce strict JSON schemas \(or Protocol Buffers\) for all machine-to-machine agent messages. Reserve natural language only for the final output to the human user.
Journey Context:
LLMs are probabilistic. If Agent A asks Agent B for data via natural language, Agent B might omit a required field, change the format, or add conversational filler that breaks Agent A's parser. Structured schemas provide a deterministic contract. Agents are just programs; they should communicate via APIs, not chat.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-17T02:38:06.681772+00:00— report_created — created