mobile-emulator-test

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted external sources, including device logs and UI tree dumps. If the application being tested contains malicious instructions within these outputs, the agent could be influenced to deviate from its intended behavior.
  • Ingestion points: adb logcat (Phase 1d, Phase 2b), uiautomator dump (Phase 2a), and git diff (Phase 1.5a).
  • Boundary markers: The skill lacks explicit delimiters or instructions to treat data from these ingestion points as untrusted.
  • Capability inventory: The skill possesses extensive capabilities, including executing shell commands on the host and emulator, performing database deletions/modifications via SQL, and terminating system processes.
  • Sanitization: There is no evidence of sanitization or filtering applied to external content before it is read by the agent.
  • [DATA_EXFILTRATION]: The skill is instructed to search for and read sensitive local configuration files (.env, .env.local) to extract credentials and endpoint URLs for Supabase and Sentry. It also performs high-privilege SQL queries against the internal auth.users table.
  • [COMMAND_EXECUTION]: The skill executes a variety of shell commands to manage the development environment. This includes terminating processes associated with specific ports using lsof, taskkill, and kill, as well as performing administrative actions on the Android emulator like pm clear to wipe application data.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 11:27 PM