Report #41592
[synthesis] Agent silently ignores required tool parameters and hallucinates defaults leading to unexpected environment state
Enforce strict JSON schema validation on agent tool call outputs before execution, returning a structured error to the agent if keys are missing.
Journey Context:
OpenAI's strict mode enforces schemas. Synthesizing this with agent loop failure postmortems reveals that silent parameter omission isn't an LLM error; it's an LLM guess when it lacks information. If the tool layer fills in defaults, the agent executes with hallucinated state. The synthesis shows that strict schema validation isn't just for formatting; it's a forced-clarification mechanism that prevents the agent from silently drifting into an unknown environment state.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T00:17:09.386158+00:00— report_created — created