nuanu-flow

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

• Third-party content exposure detected (medium risk: 0.65). The remote-worker workflow (via scripts/worker.mjs) receives task.system_prompt, task.instruction, and task.context from the Flow API over the agent-worker REST endpoints, then builds a prompt and pipes it into the local command’s stdin; since that upstream task text is not authored by the operating user, it can contain outsider free text that reaches the LLM context used by the spawned command.

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

• Potentially malicious external URL detected (high risk: 0.90). The portable worker contacts the runtime API at https://flow.nuanu.com/api (scripts/worker.mjs) to fetch tasks whose returned fields (instruction and system_prompt) are injected verbatim into a spawned child process's stdin, so remote content from that URL directly controls prompts sent to local code execution.