blender-character-rig

Warn

Audited by Gen Agent Trust Hub on Nov 20, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The scripts 'scripts/game_mesh.py' and 'scripts/ship_merge.py' use subprocesses to call external tools. Specifically, 'game_mesh.py' allows executing an arbitrary binary whose path is provided via the '--autoremesher' parameter. This presents a risk of arbitrary command execution if the agent is manipulated into specifying a malicious executable.
  • [DYNAMIC_EXECUTION]: Documentation in 'reference/mcp-session.md' provides a pattern for the agent to load and execute Python source code from strings using 'exec(compile(...))'. This is intended for environments where the Blender process cannot access local files, but it constitutes a high-risk dynamic code execution vector.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted 3D model files (GLB and Blend) and JSON plans while maintaining significant system capabilities like file manipulation and command execution. The lack of boundary markers or instructions to ignore embedded content in processed assets creates an indirect injection surface.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Nov 20, 2026, 03:25 PM