routeros-dhcp
Pass
Audited by Gen Agent Trust Hub on Oct 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill documents the handling of untrusted data from DHCP clients, which creates a potential surface for indirect prompt injection if the agent executes scripts using this data without proper sanitization.
- Ingestion points: Untrusted data enters the agent's operating context through DHCP options such as hostnames, vendor-specific information, and relay agent IDs, as detailed in
references/cli/ip/dhcp-client.mdandreferences/cli/ip/dhcp-server.md. - Boundary markers: The provided documentation and script examples for RouterOS do not include explicit delimiters or instructions to ignore embedded commands within client-provided variables.
- Capability inventory: The skill is designed to enable the agent to perform sensitive RouterOS operations, including CLI command execution, script creation, and network configuration changes.
- Sanitization: The reference examples do not demonstrate sanitization or validation of external content (e.g.,
$"vendor-specific"or$"lease-hostname") before they are interpolated into configuration commands or scripts.
Audit Metadata