Security researchers have identified a technique where malicious actors inject fake Sentry error reports to trigger unauthorized code execution by AI coding assistants. This attack forces AI agents to treat compromised diagnostic data as trusted instructions, potentially exposing sensitive environment credentials to remote attackers.