oneshot-research

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

• Third-party content exposure detected (high risk: 0.95). The required workflow uses agent.research / agent.webSearch / agent.webRead, which at runtime fetches public web pages and scraped content (outsider-authored free text) and injects it into the agent’s LLM context as markdown/report text.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill provides a runtime webRead call that fetches and injects arbitrary external pages into the agent context (e.g., https://example.com/article), so remote content can directly control prompts used by the agent.