knowledge-graph-builder

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

• Third-party content exposure detected (high risk: 0.80). The skill's documentation and pipelines explicitly admit ingesting untrusted web content (references/entity-extraction.md lists "Unstructured: Documents, web content, text files"; ontology examples include source_url and a Neo4j relationship with source 'linkedin_scrape') and those ingested documents are used to build/ground the knowledge graph that the AI Integration (references/hybrid-architecture.md and references/ai-integration.md) uses to generate and verify LLM responses, so third-party text can materially influence agent behavior.