The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes command-line tools such as adb, frida, and objection to interact with mobile devices and perform runtime exploration. It also uses standard Unix utilities like grep and strings to analyze decompiled source code and binary assets for sensitive information.
[DYNAMIC_EXECUTION]: The skill automates the creation of a Python reporting script (generate-report.py) and JavaScript instrumentation payloads (ssl-bypass.js, ios-keychain-dump.js) based on static templates provided in the reference material. These scripts are subsequently executed to aggregate test findings and bypass mobile security controls like SSL pinning.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a potential surface for indirect prompt injection as it ingests and processes untrusted data from mobile application binaries and device filesystems to generate security reports.
Ingestion points: Data is ingested from the target application binary (TARGET_APP) through static analysis tools (JADX, apktool) and from the mobile device's data directories (/data/data/ or /var/mobile/Containers/).
Boundary markers: There are no explicit delimiters or instructions provided to the agent to treat the contents of the mobile app as untrusted during the reporting phase.
Capability inventory: The skill has the ability to read and write files, execute shell commands, and perform dynamic instrumentation on external processes.
Sanitization: There is no evidence of sanitization or escaping of the metadata or strings extracted from the target application before they are included in the generated findings report.

mobile-pentest