pyexfil

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs reading sensitive files (e.g., open('secrets.txt','rb').read()) and sending their contents over covert channels, which requires the agent to handle and transmit secret values verbatim for exfiltration.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This package is explicitly designed to perform covert, multi-channel data exfiltration (including examples that send arbitrary files and microphone captures to attacker-controlled endpoints) and contains documentation about detection evasion, so it is intentionally abuse-capable and high risk.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 1.00). The skill explicitly instructs reading local files and sending them out over covert channels (data exfiltration), which actively compromises the machine’s security and enables unauthorized disclosure even if it doesn’t request sudo or change system files.