Report #51962
[architecture] Unstructured text output from Agent A breaks Agent B's input parsing
Enforce strict JSON schema contracts at agent boundaries using structured outputs or function calling, treating agent handoffs like API contracts.
Journey Context:
Agents default to free-text, and prompt engineering \('output JSON'\) is brittle. Downstream agents hallucinate parsing or fail when schemas shift. Enforcing structured output \(like OpenAI's Structured Outputs or Outlines\) guarantees schema adherence. The tradeoff is slightly restricted LLM expressiveness, but it ensures deterministic routing and prevents cascading parse failures across the chain.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T17:42:52.226308+00:00— report_created — created