iot-pentest

Pass

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and process data from external, untrusted network services such as MQTT brokers and CoAP servers. This creates an inherent surface for indirect prompt injection should the data retrieved from these services contain malicious instructions intended to manipulate the agent's logic.
  • Ingestion points: The agent ingests data from IoT protocol payloads (e.g., via mosquitto_sub) and device discovery responses (e.g., via nmap or avahi-browse) into its working context.
  • Boundary markers: No specific delimiters or "ignore previous instructions" warnings are used to frame data retrieved from target systems.
  • Capability inventory: The skill possesses the ability to execute shell commands (Bash), perform network requests (WebFetch), and modify local files (Write, Edit).
  • Sanitization: The instructions do not specify sanitization or validation of the retrieved network payloads before they are processed by the agent.
  • [EXTERNAL_DOWNLOADS]: The skill's setup instructions involve downloading legitimate security tools and libraries from well-known sources, including official package registries (PyPI) and established open-source organizations on GitHub (e.g., OWASP, Project-CHIP, Akamai). These downloads are consistent with the skill's primary purpose of providing an IoT security testing environment.
  • [COMMAND_EXECUTION]: The skill makes extensive use of the Bash tool to execute reconnaissance and exploitation commands. While these commands include powerful payloads (such as command injection templates and reverse shell examples), they are documented strictly as test cases and payloads for attacking target IoT devices in a penetration testing context, rather than the agent's host environment.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 19, 2026, 04:36 PM