reverse-engineering

Pass

Audited by Gen Agent Trust Hub on Oct 31, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill documentation suggests using uv to download and install the minilzo package from PyPI, a standard package registry, to facilitate LZO decompression during the unpacking process.
  • [DYNAMIC_EXECUTION]: The references/ue3-packages/README.md file provides a template Python snippet for users or agents to execute via shell heredoc to manually modify binary headers in processed packages.
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external game data files (e.g., .upk packages), which constitutes an attack surface for indirect prompt injection if the files are specifically crafted to exploit analysis tools.
  • Ingestion points: The unpack_upk.py script and various mentioned RE tools (Ghidra, ILSpy, etc.) read binary data from user-supplied game files.
  • Boundary markers: No specific prompt-based boundary markers are implemented in the binary processing scripts.
  • Capability inventory: The skill possesses file read/write capabilities and provides instructions for executing numerous third-party binaries and MCP servers.
  • Sanitization: The unpack_upk.py script performs structural validation by checking magic tags (0x9E2A83C1) and verifying chunk sizes before decompression.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 31, 2026, 12:40 AM