Codex auto-review

How a second model judges risky agent tool calls before they run, and how it holds up against prompt injection

Read external article →

ai agents security