The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches tools, binaries, and Android modules from various third-party GitHub and GitLab repositories.
Downloads the rootAVD toolkit from gitlab.com/newbit to automate the rooting process.
Downloads Magisk (topjohnwu/Magisk), Frida server (frida/frida), LSPosed (LSPosed/LSPosed), and several SSL pinning bypass modules (Dr-TSNG, ViRb3, Fuzion24) from GitHub.
[REMOTE_CODE_EXECUTION]: Encourages the execution of community-contributed scripts directly from the Frida CodeShare repository.
Instruction: frida -U -f com.example.app --codeshare akabe1/frida-multiple-unpinning fetches and executes code from a remote service.
[COMMAND_EXECUTION]: Instructs the user to execute shell scripts and system utilities with high-level access.
Execution of ./rootAVD.sh after cloning an external repository.
Heavy use of adb commands to push files, modify permissions (chmod 755), and interact with the system shell.
[PRIVILEGE_ESCALATION]: Provides instructions to gain administrative (root) access to the emulator environment.
Uses adb root and adb remount to grant full control over the file system.
Employs the -writable-system flag during emulator startup to allow modifications to protected partitions.
[PERSISTENCE]: Establishes persistent modifications that remain active across reboots.
Installs custom CA certificates into the system root store at /system/etc/security/cacerts/.
Installs Magisk and LSPosed modules that integrate into the boot process and zygote process of the Android OS.
[DYNAMIC_EXECUTION]: Leverages dynamic instrumentation to intercept and modify application logic at runtime.
Uses Frida to hook Java methods for bypassing security checks such as root detection (RootBeer.isRooted) and SSL pinning.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface where instructions could be embedded in the data processed by the agent.
Ingestion points: The agent processes external package names, Frida script paths, and potentially application metadata during analysis (e.g., in SKILL.md).
Boundary markers: No delimiters or safety warnings are present to distinguish instructions from the application data being analyzed.
Capability inventory: The skill possesses capabilities to execute shell commands (adb), write to the file system (adb push), and perform network operations (curl, pip).
Sanitization: There is no evidence of input validation or sanitization for application identifiers or script content before they are used in commands.

android-studio-pentest-environment