ai-integration

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

• Third-party content exposure detected (high risk: 0.80). The skill's agent/docs explicitly show agent tool patterns that fetch and use external web content (e.g., the ReAct "search(query)" tool in agents/prompting/patterns.md and the LangChain searchTool in agents/llm/langchain.md) and examples that pass image URLs to model APIs (agents/llm/claude.md and agents/llm/openai.md), meaning the agent is expected to ingest arbitrary third‑party URLs/results as part of its workflow and can feed those results back to the LLM (risking indirect prompt injection).