blender-iterative-refinement
Pass
Audited by Gen Agent Trust Hub on Aug 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data including user-visible requests and image frames, which could contain instructions intended to influence the agent's logic.
- Ingestion points: User-visible requests and preserved image files (
SKILL.md). - Boundary markers: The instructions lack explicit boundary markers or delimiters for untrusted user input.
- Capability inventory: The skill can modify durable Python/Blender source files and run the Blender executable to regenerate artifacts (
SKILL.md). - Sanitization: No specific sanitization steps are documented for handling external inputs before they are used to guide the repair process.
- [COMMAND_EXECUTION]: The workflow requires the execution of the Blender binary and the interpretation of Python source code to perform its primary function.
- Evidence: The 'Recheck and decide' section in
SKILL.mdexplicitly instructs the agent to 'Repeat the exact Blender executable' and modify 'durable Python or editable Blender source'.
Audit Metadata