Claude Code's "Don't Ask Again" feature for approving actions has been found to accidentally expose passwords and secrets in 1,122 public GitHub repositories. The bug occurs when the approval option is saved and subsequently commits sensitive credentials into publicly visible code files.
Background
APIiro published findings on security vulnerabilities associated with AI coding agents like Claude Code, Anthropic's AI programming assistant that can autonomously execute commands in development environments.
- Source
- Hacker News (RSS)
- Published
- Nov 8, 2026 at 04:45 AM
- Score
- 7.0 / 10