effect-ai-tool

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

• Third-party content exposure detected (high risk: 0.90). The references include concrete handler patterns that fetch and return arbitrary web content (references/13-common-patterns.md "FetchTool" uses fetch(url).then(r => r.text())) and a provider-defined WebSearch example (references/07-provider-defined-tools.md) whose results are passed back via toolkit.handle to the LLM, clearly exposing the agent to untrusted third-party web content that could carry indirect prompt injections.