geo-analyzer

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

• Third-party content exposure detected (medium risk: 0.30). Skill runtime path reads outsider-authored free text from output/search_results.json—specifically the content field populated from AI platform responses—then constructs prompts using that content in scripts/lib/analyzer.py (build_analysis_prompt) for Step 4 AI analysis.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill calls the REDFOX API at https://redfox.hk (API_BASE in scripts/lib/platforms.py) at runtime to submit and poll AI search tasks, and the responses from that endpoint are injected into the AI analysis prompt (build_analysis_prompt in scripts/lib/analyzer.py), so remote content can directly influence agent behavior.