effect-ai-tool

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

• Third-party content exposure detected (high risk: 1.00). The skill includes examples that fetch and ingest arbitrary public web content—e.g., the "FetchTool" handler that does fetch(url).then(r => r.text()) and the provider-defined "WebSearch" tool that returns results/snippets—so the agent would read untrusted third-party webpages/search results as part of its workflow.