style-dna

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

• Third-party content exposure detected (high risk: 0.90). The skill's required workflow (Bước 1.c "URL (Substack, blog)" in SKILL.md) explicitly fetches and crawls public Substack/blog pages via substack-tools and WebFetch and then reads/interprets that corpus to generate actionable "công thức tái tạo" style guides, so untrusted user-generated web content can directly influence the agent's analysis and subsequent outputs.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches user-provided Substack/blog URLs at runtime (via WebFetch and the referenced substack-tools/substack_crawl.py) and injects the fetched article content into the agent's context to drive prompt outputs, so these runtime-fetched URLs (user-supplied Substack/blog URLs) present a risk of remote content controlling the agent.