meta-ads-strategy

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

• Third-party content exposure detected (medium risk: 0.65). The skill’s runtime workflow can ingest outsider-authored free text when it performs Phase 1 “research proactively” by visiting the user’s website and “look up competitors” (public web pages/articles scraped/extracted at runtime), and also when it asks the user to provide “landing page URLs” whose page body text is then read into the LLM context.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill's brief.md Phase 1 explicitly tells the agent at runtime to "visit their website" and research competitor sites (i.e., fetch arbitrary external URLs such as the user-provided product/competitor pages and referenced resource https://github.com/coreyhaines31/marketingskills/blob/main/skills/lead-magnets/SKILL.md), and use that fetched content to populate the ad brief and drive subsequent prompts—so external pages directly control the agent's prompts during runtime.