5g-telecom-attack
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTIONPERSISTENCEINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads multiple open-source telecom security tools from GitHub, including Open5GS, UERANSIM, srsRAN_Project, PacketRusher, free5GC, sctpscan, and ss7MAPer. These are well-established projects in the cellular research community.
- [COMMAND_EXECUTION]: Instructions include the use of
sudofor system-level configuration necessary for telecom labs, such as loading kernel modules (modprobe sctp), managing firewall settings (ufw disable), and installing system dependencies. - [DYNAMIC_EXECUTION]: The skill uses Python heredocs to generate and execute network packet crafting scripts for protocols like PFCP and GTP-U. This is a standard method for providing reproducible proofs-of-concept in documentation.
- [PERSISTENCE]: The setup instructions include adding the
sctpmodule to/etc/modules-load.d/modules.confto ensure the protocol is available across system reboots, which is a common requirement for SCTP-based signaling. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted packet captures and tool outputs. Evidence Chain: (1) Ingestion: PCAP files and diagnostic logs (test-cases.md, payloads.md). (2) Boundary markers: Instructions advise redacting sensitive data in reports. (3) Capabilities: Subprocess calls via Bash tool and script execution across all guide files. (4) Sanitization: Not explicitly present in instructions but manual review of evidence is suggested.
Audit Metadata