google-tasks

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

• Potentially malicious external URL detected (high risk: 0.90). This skill instructs agents to call the AgentPMT MCP server (https://api.agentpmt.com/mcp/) at runtime to fetch live schema/instructions (via get_schema/get_instructions), and those fetched instructions are treated as authoritative guidance for the agent's behavior, meaning remote content can directly control prompts.