bears-workflows

SUSPICIOUS. The skill is largely aligned with its robotics/lab-automation purpose, but it has meaningful security risk from autonomous hardware control, unpinned dependencies, partially unverifiable local tooling (`balance-edge`), and experiment data plus API-key use through OpenRouter as a third-party gateway. I do not see confirmed malware or obvious credential-harvesting behavior.

The code appears to be a protocol generator rather than malware. However, unescaped configuration values are incorporated into executable Python source, creating a code-injection risk when generated protocols are run. Inputs should be validated and safely escaped or represented using a structured serialization approach.