Claude Code regularly sends each background agent’s transcript to its model
Claude Code reportedly uses 15-20% of tokens on a 3-word status line, indicating potential inefficiency.
Summary written automatically by AgentGid from the source's own text. Check the original before relying on it.