auto-paper-improvement-loop

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

• Third-party content exposure detected (high risk: 0.80). The skill optionally accepts a --style-ref which (via tools/extract_paper_style.py) can fetch/arXiv/http(s) URLs and caches a style_profile.md that is then read and used to steer the fix-implementation phase, so it clearly ingests untrusted public web/arXiv content that can influence edits.