network-tools
Pass
Audited by Gen Agent Trust Hub on Nov 21, 2026
Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONPRIVILEGE_ESCALATIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill documents a comprehensive set of Linux network utilities including
ip,netstat,ss,lsof,curl,wget, andtcpdump. While these are standard administrative tools, they allow the agent to execute significant system-level operations. - [DATA_EXFILTRATION]: Provides instructions for using
curlandwgetto interact with external web services, including examples for sending data via POST requests and using authentication headers (Bearer tokens, Basic Auth). This capability can be leveraged to move data from the local environment to external endpoints. - [PRIVILEGE_ESCALATION]: Includes commands that typically require administrative (root) privileges, such as modifying network interfaces (
ip link set), changing IP addresses (ip addr add), and performing raw packet captures (tcpdump). It also includes process termination commands likekill -9andfuser -k. - [INDIRECT_PROMPT_INJECTION]: The skill patterns rely on external inputs such as hostnames and URLs (e.g.,
ping hostname,curl http://example.com). If an agent uses this skill to process unsanitized data from a user or external source, it could be vulnerable to command injection or Server-Side Request Forgery (SSRF) attacks. - Ingestion points: Hostname and URL parameters in connectivity and HTTP tools (SKILL.md).
- Boundary markers: None present; the skill assumes direct interpolation of variables into shell commands.
- Capability inventory: Full network access, process management, and packet capture capabilities.
- Sanitization: No sanitization or validation logic is defined for the inputs used in the command templates.
Audit Metadata