Report #9967
[architecture] Router agent blindly passing tasks to specialized sub-agents without knowing if the sub-agent can actually handle the specific nuance
Require sub-agents to return a structured confidence score or validation result before the router commits to the handoff; use fallback agents for low confidence.
Journey Context:
A router might see 'code' and send to a Python agent, but the prompt is actually about C\+\+. If the Python agent just tries its best, it fails silently or hallucinates. The router needs a two-phase commit or a confidence check: 'Can you handle this?' -> 'Yes \(0.9 confidence\)' -> 'Proceed'. This prevents cascading failures down the wrong agent path.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T09:36:07.961175+00:00— report_created — created