webtoon-harness

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

• Third-party content exposure detected (high risk: 0.85). Phase 2 “Research” gathers live web data via the scrapling skill (public web extraction of rankings/comments), and that scraped free-form text is ingested into the agent context to produce 01_research/trend-brief.md.

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

• Potentially malicious external URL detected (high risk: 0.90). The research phase explicitly fetches arbitrary external pages at runtime via the scrapling skill (e.g., python -c "from scrapling.fetchers import Fetcher; print(Fetcher.get('').status)"), and those scraped source URLs feed trend-brief.md which directly controls downstream agent prompts and behavior.