marketing-sostac

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

• Third-party content exposure detected (high risk: 0.90). The skill's required Pre-Phase 0 "Auto-Discovery" (see SKILL.md and references/auto-discovery.md) explicitly runs automated browsing and scraping of public third‑party sources—Google SERPs, competitor websites, G2/Trustpilot reviews, Reddit, Quora, Meta Ad Library, Google Ads Transparency, etc.—and then pre-populates and drives the interview/workflow, so untrusted user‑generated web content is fetched, interpreted, and can materially influence subsequent actions.

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

• Potentially malicious external URL detected (high risk: 1.00). The skill's auto-discovery runs at runtime and explicitly opens external sites (e.g., agent-browser --session sostac-discovery-{brand-slug} open https://{brand-url}) and uses the fetched page content to pre-populate Phase 1 and steer the agent's interview prompts, so external URLs like https://{brand-url} are used during runtime to control agent behavior.