5g-6g-telecom-attack-advanced
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it utilizes the
WebFetchtool to retrieve data from external telecom-related domains and theBashtool to process information. An attacker could potentially embed malicious instructions in external content to influence the agent's behavior. - Ingestion points: The skill metadata specifies access to
WebFetchandReadtools for ingesting external and local data. - Boundary markers: No explicit instructions or delimiters are provided to the agent to treat external content as untrusted or to ignore embedded instructions.
- Capability inventory: The skill allows access to high-privilege tools including
Bash(shell execution),Write, andEdit(file system modification). - Sanitization: There is no evidence of input validation or sanitization logic for data retrieved from external sources.
- [COMMAND_EXECUTION]: The skill includes numerous instructions for executing shell commands and Python scripts for network scanning, protocol crafting, and lab setup.
- Evidence: Commands for
srsran,open5gs, andscapyare provided throughoutpayloads.mdandtest-cases.md. The use ofsudois instructed for package installation and starting core network services. - [EXTERNAL_DOWNLOADS]: The skill provides instructions to download and install third-party telecom security tools and core network implementations.
- Evidence: Instructions to run
sudo apt install open5gsandsudo apt install srsranare included for establishing the laboratory foundation. - [CREDENTIALS_UNSAFE]: The skill documents known default credentials for third-party software as part of its educational content on telecom vulnerabilities.
- Evidence:
payloads.mdcites CVE-2021-25863, mentioning default credentialsadmin/1423for the Open5GS WebUI. This is an instructional example for lab testing and does not represent a credential for the agent or the user's host system.
Audit Metadata