deepline-ads-audiences

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

• Third-party content exposure detected (medium risk: 0.30). The workflow ingests arbitrary user-supplied customer lists and customer data CSV files, which can contain outsider-authored free text or untrusted data payload content.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill instructs setting up and fetching from an arbitrary third-party registry (https://code.deepline.com/api/v2/npm/) and installation guide (https://code.deepline.com/INSTALL.md) which constitutes an unverifiable runtime external dependency with weak/unknown provenance, rating it high.

MEDIUM W009: Direct money access capability detected (payment gateways, crypto, banking).

• Direct money access detected (medium risk: 0.30). The skill documents operational workflows for uploading and syncing paid advertising audiences to platforms like Google Ads, Meta, and LinkedIn, as well as enriching audience datasets using paid third-party tools via Deepline CLI and plays. However, these actions upload data and query/manage audiences rather than moving, charging, or requesting money, or initiating direct financial instructions or payment settlements. The mentions of "budget" and "spend" refer to advertising spend, tool enrichment costs, and budget gates within analytics and audience coverage planning, rather than direct financial operations or wallet debits executed by the skill itself. Therefore, the financial operation is indirect (billable usage / data enrichment tracking), making the severity medium.