global-health-public-health-data

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

• Third-party content exposure detected (high risk: 0.75). The required runtime workflow calls the World Bank countries/WDI APIs to fetch indicator values; these are public web-sourced datasets (outsider-authored content) that are ingested as readable JSON/prose into the agent’s LLM context.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly instructs runtime calls to the AgentPMT MCP endpoint (https://api.agentpmt.com/mcp/) — e.g., using agentpmt-tool-search-and-execution with action=get_schema/get_instructions — and says to treat returned live schema/instructions as authoritative, so fetched remote content can directly control prompts.