mobile-security

Fail

Audited by Gen Agent Trust Hub on Sep 19, 2026

Risk Level: CRITICALEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references multiple external security tools and repositories for installation and reference. This includes automated mobile analysis frameworks like MobSF (via Docker), Flutter testing tools like reflutter, and React Native analysis tools like hbctool. Educational documentation from book.hacktricks.xyz is also cited as a core resource for mobile penetration techniques.
  • [REMOTE_CODE_EXECUTION]: The educational guides within the skill contain code snippets demonstrating remote code execution vulnerabilities. Examples include using ysoserial to generate Java deserialization payloads and Python pickle exploits that execute commands via curl | bash. These are explicitly labeled as attack examples for illustrative purposes.
  • [COMMAND_EXECUTION]: The skill facilitates the execution of numerous command-line utilities for mobile security research via the Bash tool, including adb for device interaction, apktool and jadx for decompilation, and frida for dynamic instrumentation.
  • [INDIRECT_PROMPT_INJECTION]: The skill's workflow involves analyzing external, untrusted mobile application artifacts (APKs, IPAs, and their source code). This creates a potential surface for indirect prompt injection if an analyzed file contains malicious instructions designed to exploit the agent's broad execution capabilities.
  • Ingestion points: Analysis of decompiled Java/Smali source code, AndroidManifest.xml, and application resources found in target mobile binaries.
  • Boundary markers: None implemented; the agent is instructed to perform deep audits of the provided code directly.
  • Capability inventory: The skill uses tools with capabilities for file system modification (Write, Edit), shell execution (Bash), and network access (WebFetch, WebSearch).
  • Sanitization: There is no specialized sanitization or filtering logic applied to the content of the mobile applications being analyzed.
  • [DYNAMIC_EXECUTION]: A core component of the skill is the use of Frida and Objection for dynamic instrumentation. This involves the runtime injection of JavaScript hooks into target applications to intercept credentials, bypass security checks like root detection and SSL pinning, and analyze encrypted traffic.
Recommendations
  • Contains 6 malicious URL(s) - DO NOT USE
Audit Metadata
Risk Level
CRITICAL
Analyzed
Sep 19, 2026, 04:36 PM