numerical-methods-expert

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

• Third-party content exposure detected (high risk: 0.75). The required workflow is a “fresh web research” step (targeted web searches and recording sources) which, at runtime, can fetch public web pages and ingest their scraped free text into the agent’s LLM context (public web content category).

MEDIUM W021: Hidden or invisible Unicode characters detected (potential obfuscation or prompt injection).

• Hidden Unicode characters detected (1 type(s) found)