sales-leedlime

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

• Third-party content exposure detected (high risk: 0.75). Outsider free text can enter the LLM context via the runtime “Reddit monitoring” path: Leedlime surfaces Reddit posts/comments authored by other users, and the skill routes questions to platform guidance while the agent may use those lead/post texts to answer or draft replies (public outsider-authored content).

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

• Potentially malicious external URL detected (high risk: 0.85). The skill explicitly ingests user-supplied external website/competitor URLs at runtime (e.g., "Website URL" and "Competitor URLs (up to 3)") which the platform "uses for AI to understand your product" and positioning context, so fetched webpage content can directly influence the agent's prompts and scoring (i.e., user-provided product/competitor URLs).