The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data, including Android APKs, iOS IPAs, and intercepted network traffic.
Ingestion points: The agent analyzes external files such as Manifests, Info.plist files, and binary symbols using tools like androguard, class-dump, and APKLeaks (e.g., in Phase 1 and Phase 2).
Boundary markers: A 'Mandatory confirmation gate' is included in SKILL.md requiring explicit user verification of targets and scope, which provides a layer of protection against unauthorized actions but does not sanitize data content.
Capability inventory: The skill possesses extensive capabilities including binary patching (apktool), runtime instrumentation (frida), and local file system modifications across multiple scripts.
Sanitization: No specific sanitization or filtering logic is defined for the content extracted from target binaries or network captures before it is presented to the agent context.
[COMMAND_EXECUTION]: The workflow involves executing numerous external CLI utilities for binary manipulation and system interaction, such as apktool, ipatool, codesign, otool, jtool2, lipo, and nm.
[DYNAMIC_EXECUTION]: The primary function of the skill is dynamic instrumentation and runtime code injection.
Frida/Objection: Extensive use of Frida scripts and Objection commands to hook Java/Objective-C methods, modify application logic at runtime, and dump sensitive data like Keystore/Keychain contents.
Binary Patching: The skill provides procedures for injecting frida-gadget into APK/IPA binaries and re-signing them to bypass security controls on standard devices.

mobile-reverse