atai-operational-state-monitoring-agent

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). Skill runtime ingests the user-provided sensor CSV content via the Files API upload and then runs the pre-packaged “OSM Quick Start” bundle on that data (POST {endpoint}/v0.5/files, then POST {endpoint}/agents/bundles/{id}/run with connectors.source.file.id).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill's "Bring your own classifier" bundle creation requires an s3 artifact URI that the platform will fetch and execute (s3:////my-classifier.safetensors), which is a runtime external dependency that supplies executable model weights used to control classification behavior.