Summary

• Both reports involve compromised Python packages distributed through PyPI and executing malicious code when imported.
• The payloads target high-value secrets such as cloud credentials, API keys, SSH keys, tokens, environment files, and Kubernetes secrets.
• One case added startup persistence through a .pth file, showing a move from one-time execution to durable compromise.
• Exfiltration is described as encrypted and directed to attacker-controlled infrastructure, indicating deliberate credential harvesting rather than opportunistic misuse.
• Affected projects were removed or rolled back, with remediation centered on removing bad releases and rotating exposed credentials.
• The pattern points to continued supply-chain risk in Python tooling, especially for packages used in infrastructure and LLM workflows.