Report #66661
[architecture] Orchestrator calling the wrong agent's tool due to identical function names across different agent skill sets
Namespace all tool names with the agent's identifier \(e.g., python\_exec vs bash\_exec\) and strictly scope tool availability per agent turn.
Journey Context:
In a flat tool registry, name collisions cause unpredictable behavior. If the 'Database Agent' and 'API Agent' both expose a run\_query tool, the orchestrator will unpredictably invoke the wrong one. Namespacing ensures global uniqueness. Furthermore, passing all tools from all agents to a single orchestrator overwhelms the LLM's attention mechanism and increases hallucination. Scoping tools—only providing the tools available to the currently active agent—drastically reduces prompt size and improves routing accuracy.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-20T18:22:28.726956+00:00— report_created — created