stop-hook-feedback-loop-bloats-context-window

Each false-positive Stop hook invocation injects a large feedback block into the conversation as a user message, compounding context consumption across the session. In this session, ~30 identical hook feedback blocks were injected, each consuming significant tokens and degrading conversation quality. The hook must be scoped to fire only on confirmed agent stop events, not idle/waiting states.

Related

• stop-hook-false-positive-on-minimal-agent-responses
• stop-hook-false-positive-on-waiting-state
• stop-hook-false-positive-loop-on-waiting-state
• stop-hook-false-positive-during-active-waiting-state
• stop-hook-fires-on-every-turn-not-just-agent-stop