Autodesk
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONMETADATA_POISONING
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends installing the
fluepackage from PyPI (https://pypi.org/project/flue) and references its GitHub repository (https://github.com/SFKislev/flue), both managed by the skill author. - [DYNAMIC_EXECUTION]: The core mechanism involves generating and running scripts (MAXScript, ExtendScript, etc.) inside desktop applications, which is a powerful dynamic execution surface.
- [COMMAND_EXECUTION]: The agent uses shell commands to pipe scripts into local bridge processes like
3dsmax_bridge.py. - [INDIRECT_PROMPT_INJECTION]: The skill acts as a proxy for user commands into desktop software, creating a vulnerability where malicious input could lead to unauthorized application actions. Mandatory Evidence Chain: 1. Ingestion points: User instructions processed by the agent; 2. Boundary markers: None; 3. Capability inventory: Bridge scripts and application script runtimes; 4. Sanitization: Safety instructions emphasizing human oversight.
- [METADATA_POISONING]: The skill contains a hardcoded environment path (C:\Users\fredd.claude\skills\flue\SKILL.md) that may leak author context or cause issues on systems with different structures.
Audit Metadata