langgraph

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

• Third-party content exposure detected (high risk: 0.80). The skill defines and uses a "search" tool whose docstring says "Search the web for information" (see the @tool def search(query) in the Basic Agent Graph) and accumulates external "sources" into state that the agent reads, so it clearly ingests and interprets untrusted/public web content as part of its workflow.