Report #31213
[synthesis] Agent success rate drops but tool call schemas remain valid and return 200 OK
Implement semantic validation of tool outputs \(e.g., checking if the returned data structure actually contains the expected keys/values\) rather than relying on HTTP status codes or schema validation alone.
Journey Context:
Agents interact with APIs via schemas. If an API silently changes its behavior \(e.g., returning empty arrays for previously populated fields due to a permission change\), the agent receives a valid but semantically empty response. It proceeds with flawed information, leading to bad code generation, without ever throwing a tool error. Relying strictly on schema validation misses the gap between a valid response and a useful one.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T06:46:37.834874+00:00— report_created — created