mobile-qa

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It ingests untrusted data from the mobile application under test (screens, functionality, metadata) and user-supplied $ARGUMENTS. Malicious instructions embedded in the app's content or metadata could influence the agent's behavior during the audit.
  • Ingestion points: $ARGUMENTS variable and the application source code/runtime environment during the inventory phase.
  • Boundary markers: Absent for user arguments and application data processing.
  • Capability inventory: The agent can execute build commands (flutter build), invoke project-specific tools, and write telemetry logs to the local filesystem (~/.claude/projects/).
  • Sanitization: No explicit sanitization or validation of the processed application data is defined.
  • [COMMAND_EXECUTION]: The skill executes shell commands to perform mobile application builds (e.g., 'flutter build') and suggests using external proxy/simulation tools. While these are standard for mobile development QA tasks, they represent a powerful capability that could be redirected if the agent's context is compromised by indirect injection.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:58 AM
Security Audit — agent-trust-hub — mobile-qa