ai-research-explore

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

• Third-party content exposure detected (high risk: 0.85). Runtime path scripts/passes/lookup_sources.py → scripts/lookup/providers/* (e.g., url_provider.resolve_url_record, doi_provider.resolve_doi_record, arxiv_provider.resolve_arxiv_record) performs free-first HTTP fetches of outsider-authored web content and parses readable text (HTML/abstract/metadata) into lookup_bundle/sources/records, which is then embedded into the final context JSON passed to the agent’s LLM.