Report #1798
[architecture] Free-form natural language messages between agents cause hallucinated parsing and coordination failures
Enforce structured-schema messaging \(e.g., JSON Schema, Pydantic models\) for all inter-agent communication, restricting natural language to user-facing outputs only.
Journey Context:
It is tempting to let agents talk to each other via unstructured text because LLMs excel at language. However, the receiving agent often misinterprets intent, misses parameters, or hallucinates function calls based on prose. By forcing a strict schema for machine-to-machine communication, you guarantee the contract. OpenAI's structured outputs and function calling APIs were built specifically to solve this unreliability in agent tool-use and routing.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-15T08:30:54.077598+00:00— report_created — created