loki-mode

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 0.90). The content presents a high-risk autonomous execution framework that explicitly instructs agents to bypass permissions (--dangerously-skip-permissions), run commands/commits/deployments without human approval, and can enable prompt injection (LOKI_PROMPT_INJECTION), creating strong potential for data exfiltration, remote code execution, credential theft, and supply-chain compromise even though no single line of obfuscated payload or hardcoded exfiltration URL is present.

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 requires a --dangerously-skip-permissions flag and instructs fully autonomous execution of commands, commits, deployments, and file writes without human confirmation, effectively directing the agent to bypass permission checks and modify the host state.