Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

WeaknessASC-2026-0038Other weakness

Auto-approve and unattended modes for agents

MediumMitigated

Modes that skip confirmation (auto-run, yolo, dangerously skip permissions) let any injected instruction run commands and change files without a person seeing it.

How to detect it

  • Agent settings that disable approval prompts
  • Agents running headless in CI with broad tools

How to resolve it

  • Use auto-approve only inside disposable sandboxes without secrets
  • Keep allowlists explicit and narrow
  • Require approval for network, write and deploy actions

How to investigate

  1. Find where approval was disabled and what ran without review
  2. Review the session transcript for injected instructions

Sources

Open it in the Learning Center's vulnerability database