The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill operates by ingesting and analyzing untrusted iOS app binaries, manifests, and strings, which could contain malicious payloads designed to subvert the analysis process.\n
Ingestion points: Analysis of target.ipa, manifest.plist, and GoogleService-Info.plist (SKILL.md).\n
Boundary markers: Includes a mandatory confirmation gate requiring the user to explicitly confirm research scope and authorization.\n
Capability inventory: Requires multiple local execution capabilities including unzip, grep, plutil, and the frida instrumentation framework.\n
Sanitization: Recommends the use of isolated environments, such as dedicated test devices and virtual machines.\n- [EXTERNAL_DOWNLOADS]: The skill documents the installation of necessary security utilities.\n
Evidence: Instructions to install libimobiledevice, ideviceinstaller, class-dump, frida-tools, objection, and iphone_backup_decrypt (SKILL.md).\n
Note: These are established and well-known tools in the security community.\n- [COMMAND_EXECUTION]: The skill utilizes a large number of system commands to perform its stated function.\n
Evidence: Shell commands using curl, grep, find, and plutil are provided throughout the document (SKILL.md).\n
Security Context: A flagged pipe to python3 was analyzed and found to be a safe use of the standard library json.tool for data formatting rather than remote code execution.

ios-redteam-pipeline