spy-ads

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md ingests outsider-authored free text at runtime from Facebook Ad Library pages via the Apify actor/browser scrape (Stage 1) and then pulls transcripts/copy/body text from the scraped ads for analysis (Stage 3), without requiring a prior selection beyond the user’s hunt/shortlist flow.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls Arcads' hosted MCP (https://mcp.arcads.ai) at runtime via arcads_analyze_media to produce a "reconstructed prompt" that is returned and used in teardowns/clone flows, so remote content directly supplies prompts the agent consumes.