Agent fails: Dots subagent refuses user approval passed via main agent
xeophon · x · 2026-10-05
Grant Slatton hit an annoying failure mode while using Dots to clean up unused AWS resources: the task was delegated to a subagent, which refused to accept his explicit approval relayed through the main agent. xeophon adds a comparison: Dots mostly proactively sends stale Slack messages he already knew about or replied to, while Muse's proactive follow-ups (reminders, related findings) have been genuinely useful. Together the thread highlights two immature spots in current agents: proactive triggering quality, and how human approvals propagate across multi-agent delegation chains.
More from coding & agent
- Still running Claude Code in ask-permission mode? That's a trust issue — jaivinwylde · 2026-10-05
- He sold a garage of old GPUs with ChatGPT Computer Use running his entire eBay workflow — doodlestein · 2026-10-05
- 5 projects put Jev to work as the decision engine inside agents — Arindam_1729 · 2026-10-05
- Dev ditches AI code reviewers: external reviewers redundant, invest tokens in automated QA instead — _AustinCalvert_ · 2026-10-05
- Agent coding's new bottleneck is decision-making, and chat UIs can't keep up — _AustinCalvert_ · 2026-10-05
- Microsoft's Agensh: boss-free coding agent teams keep improving from 1 to 128 agents — rohanpaul_ai · 2026-10-05