Report #4230
[architecture] Creating a new agent when a simple API tool call would suffice
If a sub-task requires no independent reasoning, system prompt, or state tracking, implement it as a tool/function on an existing agent, not as a separate agent.
Journey Context:
A common anti-pattern is wrapping every API call or utility in its own agent. Agents are autonomous loops with their own system prompts and context windows. If the sub-task is deterministic or requires no multi-step reasoning \(e.g., 'fetch weather', 'read file'\), an agent wrapper is pure overhead. Agents are for reasoning; tools are for doing.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-15T19:03:53.189966+00:00— report_created — created