create-farmer

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). The skill explicitly enables automated harvesting and scheduled subagent execution with default full-environment access and auto-push behavior, creating clear avenues for data exfiltration and persistent backdoor-like operation.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow has the agent discover and probe an MCP service (via ToolSearch and read-only tool tests) and then read “source (24hr max)” data from that external service into the farmer’s system prompt/body, which is outsider-authored free text at runtime (public/third-party content being ingested and used as LLM context).