covert-network-triggers

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The document describes a covert multi-protocol trigger/backdoor: it hides and decodes payloads inside normal network protocols, validates an implant key, then forks a child to initiate a callback to an external IP/port (C2), using raw sockets and obfuscation—behavior clearly intended for covert command-and-control and remote activation/exfiltration.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.70). The skill is a covert network trigger system that ingests attacker-controlled network packet fields (e.g., DNS subdomains, TFTP WRQ filenames, ICMP timestamp/header fields, raw TCP/UDP payload bytes) at runtime; those outsider-authored free-text values can be parsed into LLM context if any part of the workflow forwards extracted packet text to the agent’s LLM (indirect prompt injection 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 prompt instructs building a covert network implant/backdoor that listens for crafted triggers and forks callbacks, explicitly requires raw sockets/IPPROTO_RAW (privileged operations) and creates unauthorized persistent network behavior—effectively enabling remote control of the host.