The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is explicitly designed to collect and parse sensitive system telemetry exfiltrated from remote "implants." The protocol handling logic extracts MAC addresses, process lists, network configurations (ifconfig), and active network connections (netstat), which are standard targets for data exfiltration during cyber-attacks.- [REMOTE_CODE_EXECUTION]: The skill provides the server-side logic for an "implant handler" and "beacon gateway." These are core components of Command and Control (C2) infrastructure used to manage remote code execution on compromised machines.- [PROMPT_INJECTION]: The skill ingests unauthenticated, encrypted binary data from a network socket and parses it into structured data for report generation. This processes external, attacker-controlled data without sanitization, creating a significant indirect prompt injection attack surface where malicious instructions could be embedded in beacon survey fields (like process names or network labels) to manipulate the AI agent's subsequent analysis.
Ingestion points: The listen_for_beacons and process_ver2_beacon functions receive external data via conn.recv() (SKILL.md).
Boundary markers: None identified; the payload is processed directly based on BTHP protocol headers.
Capability inventory: The skill can open network ports, communicate over TCP, and write XML reports to the local system.
Sanitization: There is no validation or sanitization of the decrypted beacon data before it is used to populate XML reports.

beacon-management-gateway