mobile-recon-android
Pass
Audited by Gen Agent Trust Hub on Jul 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the installation of gplaycli, frida-tools, and objection through the pip package manager. These packages are not pinned to specific versions and are retrieved from external repositories during runtime.
- [COMMAND_EXECUTION]: The skill makes extensive use of system commands including adb, apktool, jadx, and docker to perform analysis and interact with mobile devices.
- [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface due to the processing of untrusted APK files. * Ingestion points: The agent reads and processes binary APK files and their decompiled contents. * Boundary markers: No delimiters or warnings are used to differentiate untrusted data from instructions. * Capability inventory: The skill utilizes adb for device interaction, curl for network requests, and multiple file-processing tools. * Sanitization: Extracted strings and identifiers are used directly in shell commands and network requests without verification or escaping.
- [DATA_EXFILTRATION]: The skill performs automated network requests to Firebase database endpoints extracted from the analyzed application to verify read and write permissions.
Audit Metadata