obey

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). This skill requires the agent to copy and execute the user's command verbatim, which would force the LLM to reproduce any embedded secrets (API keys, tokens, passwords, cookies) exactly, creating a direct exfiltration risk.

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document explicitly instructs executing user-supplied commands verbatim with "no exceptions", deliberately enabling remote code execution and making abuse (malicious command execution, data exfiltration, credential theft) trivially possible.

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

• Attempt to modify system services in skill instructions detected (high risk: 1.00). This skill instructs the agent to execute the user's command verbatim with "no exceptions", forcing it to run arbitrary shell commands (including privileged or state-changing ones like sudo, user creation, or system file edits) and thereby risks compromising the machine state.