firebase-apk-scanner
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The workflow instructions in SKILL.md use the '$ARGUMENTS' variable without quotes in shell commands (e.g., 'ls -la $ARGUMENTS' and '{baseDir}/scanner.sh $ARGUMENTS'). While many agent platforms sanitize tool arguments, this pattern theoretically allows a user to provide an APK path containing shell metacharacters or command separators (e.g., '; rm -rf /') to execute arbitrary commands through the agent's Bash tool.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted APK files provided by the user and performs complex string extraction and network operations based on the content. The ingestion of binary data which is then parsed by shell utilities (grep, sed, awk) to formulate network requests creates an attack surface for indirect injection, though the severity is low as the scanner script generally quotes variables.
Audit Metadata