re-mobile
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted mobile application files (APK, IPA, SO, DYLIB) which could contain adversarial content in metadata or strings intended to influence the agent's analysis.
- Ingestion points: Files are ingested in step 1 and 2 for triage and static analysis (SKILL.md).
- Boundary markers: None explicitly defined for file content boundaries, though sandboxing is recommended for dynamic execution.
- Capability inventory: Includes dex-parsing, JNI analysis, and Frida instrumentation (SKILL.md).
- Sanitization: The skill recommends network isolation and simulator snapshots for dynamic analysis to mitigate risks from untrusted code execution.
- [COMMAND_EXECUTION]: The workflow involves the use of standard command-line tools for reverse engineering, such as apktool, jadx, class-dump, and lldb.
- [DYNAMIC_EXECUTION]: The skill utilizes Frida for runtime instrumentation and memory dumping, which involves executing scripts within target application processes. This is a primary and documented function of the skill for security research.
Audit Metadata