If you're spinning up background agents in Claude, stop polling for their output. The system already pushes reminders when agents make progress or write files, so there's no need to waste tokens on repeated checks or flood your context with TaskOutput calls. Launch your agent with a specified output path, move on to other work, and wait for the system notification. Then read the output file directly. The pattern is simple: trust the built-in progress notifications instead of burning cycles on sleep loops and tail commands that accomplish nothing.
npx skills add https://github.com/parcadei/continuous-claude-v3 --skill background-agent-pings