home-assistant

Fail

Audited by Gen Agent Trust Hub on Apr 10, 2026

Risk Level: HIGHCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill instructs the agent to access highly sensitive files on remote Home Assistant instances. Specifically, it provides examples of reading 'secrets.yaml' and 'configuration.yaml', which often contain plain-text passwords, API tokens, and private keys. Even though the description claims to check keys only, the 'cat' command reveals the entire file content.\n- [PRIVILEGE_ESCALATION]: The instructions direct the agent to use 'sudo' within SSH commands to bypass standard file system permissions when accessing configuration files or Docker logs on the remote Raspberry Pi instance.\n- [DATA_EXPOSURE_AND_EXFILTRATION]: Detailed internal network information is exposed within the skill body, including local IP addresses (e.g., 192.168.8.103, 192.168.1.8), hostname patterns, and tracking data for named individuals, which could be used for further network mapping or targeted attacks.\n- [COMMAND_EXECUTION]: The skill relies on the execution of local Python scripts and remote SSH commands to perform its functions. The ability to run arbitrary shell commands via SSH represents a significant capability that could be misused if provided with malicious input.\n- [INDIRECT_PROMPT_INJECTION]: The skill defines a large surface for indirect prompt injection by consuming untrusted data from Home Assistant logs and entity states without implementing sanitization or boundary markers.\n
  • Ingestion points: The skill reads sensor states, lists entities, and tails system logs from Home Assistant deployments (SKILL.md).\n
  • Boundary markers: Absent. There are no instructions provided to delimit or ignore instructions that might be embedded within retrieved device names, sensor values, or log entries.\n
  • Capability inventory: The skill has the ability to execute shell commands locally and remotely via SSH, as well as call Home Assistant services (SKILL.md).\n
  • Sanitization: Absent. No validation or escaping of external content is described before processing or reporting retrieved data.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Apr 10, 2026, 12:22 AM
Security Audit — agent-trust-hub — home-assistant