OpenAI Agents SDK Development

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

• Third-party content exposure detected (high risk: 0.80). The skill includes function tool examples that perform HTTP fetches of external endpoints (e.g., references/tools.md parallel_fetch uses client.get(url) for arbitrary URLs and the SKILL.md async fetch_user_data example uses httpx to GET external APIs), meaning the agent can ingest untrusted public web content which could influence its tool use and decisions.