The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous commands for interacting with Android devices via adb, frida, and objection. These include installing and running server-side components (frida-server), launching activities with custom data, and querying content providers.
[DATA_EXFILTRATION]: Includes instructions and scripts to extract sensitive application information such as SharedPreferences, SQLite databases, and system clipboard data. These methods are documented for assessing data storage security during mobile pentests.
[DYNAMIC_EXECUTION]: Heavily utilizes the Frida instrumentation framework to hook Java classes and native methods at runtime. This allows for the dynamic modification of application logic to bypass security checks like SSL certificate validation and root integrity tests.
[INDIRECT_PROMPT_INJECTION]: The skill interacts with untrusted data from Android Inter-Process Communication (IPC) mechanisms, WebViews, and command outputs, creating a potential attack surface for indirect prompt injection.
Ingestion points: Data obtained via content provider queries, broadcast intents, and external URLs loaded into WebViews.
Boundary markers: No specific delimiters or "ignore" instructions are present to prevent the agent from following directions embedded within the analyzed application data.
Capability inventory: The skill possesses capabilities for shell command execution (adb), dynamic instrumentation (frida), and file system access.
Sanitization: The provided scripts and methodologies do not implement explicit sanitization or validation for the data ingested from external application components.

android-pentesting-tricks