verify-goal

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

• Third-party content exposure detected (high risk: 0.75). The required workflow runs an Exa web_search_exa/web_fetch_exa discovery pass (public web content fetched at runtime), and any fetched page text can be ingested into the LLM context during research, creating an indirect prompt-injection path.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill declares and uses an external MCP endpoint (https://mcp.exa.ai/mcp) as a runtime tool dependency (agents/openai.yaml) and references it in installation/usage commands (references/tool-gap-analysis.md), meaning the agent will contact that external service at runtime to discover/drive verifiers and thus external content can influence agent behavior; therefore it is flagged.