product-feed-optimizer

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

• Third-party content exposure detected (high risk: 0.85). The skill’s REQUIRED workflow ingests the user-provided product feed export and diagnostics as “untrusted input,” and those files are free-form text (e.g., product titles/descriptions and landing-page text) authored by outsiders (publishers/merchants), which the agent reads into LLM context for rewriting/triage.