Feature request: credit refund or budget tracking for wasteful subagent patterns
Status Open
Maintainer reply None cached
Activity 0 comments · opened Aug 8, 2026
Problem
When Claude Code spawns subagents and then resumes them unnecessarily via SendMessage, the extra work is wasted tokens with no value to the user. In this case a subagent was continued to do deep binary analysis of the claude.exe binary after the root cause was already identified and the fix was known, generating significant noise on screen and burning credits with no benefit.
Request
- A mechanism to track token spend per agent/subagent turn so wasted credit is visible to the user.
- Optionally: a way to flag or refund credits when a subagent run produced no net value (e.g. duplicate analysis, unnecessary continuation after goal was met).
This is a UX and cost-efficiency issue - users paying for API credits have no recourse when the model makes poor orchestration decisions that burn tokens.