Indirect Prompt Injection in GitHub Copilot via Malicious Documentation Leads to Credential Exfiltration | AI Breaking Wire