The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions involve the use of standard penetration testing and reverse engineering tools such as apktool, jadx, frida, objection, adb, node, grep, and rg to analyze application binaries and perform dynamic instrumentation.
[DYNAMIC_EXECUTION]: The skill describes a methodology for deobfuscating UniApp logic by extracting JavaScript components (such as RC4 decoding functions and large string arrays) and executing the resulting assembled script using Node.js to retrieve plaintext configuration and strings.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves ingesting and processing data from external, untrusted sources including binary files (.so libraries), obfuscated JavaScript (app-service.js), and Android system files.
Ingestion points: Data is ingested from target binaries, specifically libapp.so, app-service.js, and Android Manifest files during the reversing process.
Boundary markers: None identified in the instructions.
Capability inventory: The skill uses node to execute extracted logic, frida for runtime instrumentation, and adb for device interaction.
Sanitization: None described; the agent is instructed to directly grep and extract strings from these untrusted sources. This creates a vulnerability surface where content extracted from a target binary could potentially contain instructions aimed at influencing the agent, though this is a standard risk for reverse engineering tasks.

binary-mobile-reversing