The 'magique-ai' PyPI package versions 0.4.4 and 0.4.5 were compromised to include an obfuscated JavaScript infostealer executed with the Bun runtime during Python startup. This malware collects a wide range of sensitive information including environment variables, cloud tokens, credentials, cryptocurrency assets, and password manager data. It actively queries online services to gather additional secrets and attempts to maintain persistence and propagate by publishing infected packages using stolen credentials. Data exfiltration is likely performed via GitHub. The malicious code also contains destructive capabilities that threaten data wiping if invalid GitHub tokens are detected. The campaign is associated with the Mini Shai Hulud threat group and exhibits advanced evasion techniques such as sandbox detection.

Sensitive information including API keys, credentials for package repositories, cryptocurrency assets, and password manager data can be stolen. The malware can spread further by publishing infected packages using stolen credentials, increasing the risk of supply chain compromise. There is also a risk of data destruction if invalid GitHub tokens are detected, potentially causing data loss. This compromises confidentiality, integrity, and availability of affected systems and data.

No official patch or remediation is currently available. Users should immediately uninstall versions 0.4.4 and 0.4.5 of the 'magique-ai' package and avoid using these versions. Exercise caution during cleanup due to the destructive capabilities of the malware. Monitor for any unauthorized package publishing activity using your credentials. Check vendor or PyPI advisories for updates or official guidance. Since this is not a cloud service, remediation depends on user action.