exa-contents

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

• Third-party content exposure detected (high risk: 0.85). The required workflow described in SKILL.md is to call Exa’s /contents endpoint with user-provided urls, and Exa returns extracted results[].text / results[].highlights / results[].summary (LLM-ready prose from those fetched web pages), which can contain outsider-authored free text (e.g., arbitrary public web content) that the agent would read into its LLM context.