literature-engineer

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

• Third-party content exposure detected (low risk: 0.10). The workflow fetches papers from arXiv API and Semantic Scholar using search queries, and optionally ingests local offline export files; while web search results are external, they require active search/fetch rather than persistent monitoring or unprompted inbound submission.

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

• Potentially malicious external URL detected (high risk: 0.70). The script fetches data from Semantic Scholar (https://api.semanticscholar.org/graph/v1/paper/search) and falls back to a Jina proxy (https://r.jina.ai/), which constitutes an unauthenticated third-party proxy dependency.