blender-3d

Warn

Audited by Gen Agent Trust Hub on Apr 6, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONREMOTE_CODE_EXECUTION
Full Analysis
  • [REMOTE_CODE_EXECUTION]: The skill provides a helper function 'bake_and_replace_driver' that uses the 'exec()' function to dynamically assign values to object properties during animation baking.
  • Evidence: The line 'exec(f"target.{data_path} = {values[i]}")' in 'SKILL.md'.
  • Risk: Executing dynamically generated strings is a common code injection vector. If the 'data_path' or the resolved values are influenced by untrusted external data, it could lead to arbitrary code execution within the Blender process.
  • [COMMAND_EXECUTION]: The skill instructs the agent to use shell commands to communicate with a local Blender instance.
  • Evidence: The use of 'curl -s localhost:5656 --data-binary @- <<'PYEOF'' in 'SKILL.md'.
  • Context: While the target is 'localhost', directing the agent to execute shell commands for API interaction increases the attack surface and allows for potential command manipulation.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 6, 2026, 04:19 AM
Security Audit — agent-trust-hub — blender-3d