The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation instructs the agent to download and install a repository from an external GitHub source (https://github.com/DragonJAR/Android-Pentesting-Skill) into the local agent skills directory.\n- [REMOTE_CODE_EXECUTION]: The skill's primary functions rely on executing a variety of shell and Python scripts sourced from the external repository, such as auto-audit-static.sh, runtime-defense-analyzer.sh, and frida-exploit-helper.py.\n- [COMMAND_EXECUTION]: The skill executes multiple system commands including adb, apktool, jadx, and apksigner. It also includes troubleshooting instructions that use su -c for root-level command execution on Android devices.\n- [DYNAMIC_EXECUTION]: The skill uses Frida instrumentation to perform dynamic code injection into Android processes. This technique is used to bypass SSL pinning, root detection, and other runtime protections by hooking into the application's memory space.\n- [INDIRECT_PROMPT_INJECTION]: The skill acts as an ingestion point for untrusted external data in the form of Android APK files.\n
Ingestion points: The audit scripts process user-provided APK files, extracting strings, manifest data, and decompiled code (e.g., in scripts/auto-audit-static.sh).\n
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat the content of the analyzed files as potentially hostile or to ignore embedded instructions.\n
Capability inventory: The skill possesses extensive capabilities, including executing local scripts, interacting with the file system, and manipulating running processes via Frida.\n
Sanitization: There is no mention of filtering or sanitizing the data extracted from the target APKs before it is interpreted by the agent.

dragonjar-android-pentesting-skill