Report #6973
[architecture] Multiple agents sharing a tool namespace causes the LLM to call the wrong agent's tool
Namespace all tools with agent-specific prefixes \(e.g., python\_exec\_run, bash\_exec\_run\) before injecting them into the agent's context.
Journey Context:
When orchestrators merge tools from multiple agents into a single LLM call, the LLM struggles to disambiguate tools with similar names or descriptions \(e.g., two different 'search' tools\). This leads to cross-agent tool hallucination. Namespacing forces the LLM to explicitly select the correct agent's implementation.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T01:34:35.842483+00:00— report_created — created