Report #49772
[synthesis] Agent halts prematurely on schema-valid but semantically empty API responses \(200 OK with empty arrays\)
Differentiate between 'successful completion' and 'empty success' in tool definitions. Instrument the agent to log empty data structures \(e.g., response.data.length == 0\) as a warning state requiring explicit verification before halting.
Journey Context:
Agents are trained to stop when they have the answer. If a tool returns an empty array, the schema is valid, the HTTP status is 200, and the agent often concludes 'no data exists' and returns to the user. But in production, an empty array might mean an upstream dependency is down or a query parameter was malformed. The agent sees success; the user sees a bad answer. You must teach the agent that empty data is an ambiguous state, not a terminal success state. This synthesizes OpenAPI response patterns with LLM halting conditions: the agent needs an explicit 'empty state handling' protocol in its system prompt to prevent silent false negatives.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T14:01:30.181978+00:00— report_created — created