netbox-discovery
Pass
Audited by Gen Agent Trust Hub on Oct 16, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill describes a configuration that ingests data from external, potentially untrusted sources such as network devices and remote Git repositories, which could be used as an attack vector for indirect prompt injection.
- Ingestion points: Network devices (via SNMP, NAPALM, and gNMI protocols) and policy configuration files fetched from remote Git repositories.
- Boundary markers: No specific delimiters or boundary markers for isolating untrusted discovery data from the agent's prompt context are mentioned.
- Capability inventory: The skill details capabilities including system command execution (NMAP scans), network communication (Diode API ingestion), and local file system writes (dry run output).
- Sanitization: The agent configuration supports a
sanitize_configoption to redact sensitive information such as secrets from captured device configurations. - [DYNAMIC_EXECUTION]: The skill supports the execution of custom Python packages (Workers) and NAPALM drivers provided by the user via local file paths or environment variables, allowing for runtime extension of the agent's functionality.
Audit Metadata