Blender
Pass
Audited by Gen Agent Trust Hub on May 31, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill recommends installing the 'flue' package via pip from PyPI or GitHub. It strictly instructs the agent to seek explicit human approval before proceeding with installation.
- [COMMAND_EXECUTION]: The skill functions by executing shell commands that pipe scripts into application-specific bridge processes, such as 'blender_bridge.py'. It references a local path 'C:\Users\fredd.claude\skills\flue\SKILL.md' as an operational source of truth.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its interaction with external application data.
- Ingestion points: Data from Blender scenes and desktop documents (Photoshop, Excel, etc.) are read via the bridge into the agent context.
- Boundary markers: No specific delimiters or safety instructions for ignoring embedded content are provided in the skill instructions.
- Capability inventory: The skill can execute arbitrary scripts within the host applications' scripting runtimes (e.g., Python/bpy, ExtendScript, Unity Editor API).
- Sanitization: No explicit validation, escaping, or filtering of ingested application data is documented.
Audit Metadata