fully-kiosk

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface. Ingestion points: The agent processes device status JSON from the 'deviceInfo' endpoint and MQTT event topics as described in SKILL.md and reference.md. Boundary markers: The skill does not implement delimiters or explicit instructions to disregard natural language commands that might be present in ingested device metadata. Capability inventory: The skill allows the agent to execute a wide variety of commands via curl, including browser navigation, application management, and system-level operations. Sanitization: No input validation or sanitization is mentioned for data received from managed tablets.
  • [COMMAND_EXECUTION]: The skill documents the use of curl to send management commands to tablets over the local network via the REST API.
  • [EXTERNAL_DOWNLOADS]: The skill documents APIs for downloading and installing external files, such as APKs and ZIP archives, onto managed devices (loadApkFile, loadZipFile).
  • [REMOTE_CODE_EXECUTION]: The skill documents the capability to execute root and shell commands (runRootCommand, runSuCommand) on tablets that have been specifically configured for such access.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 11:09 AM