nemo-retriever

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

• Third-party content exposure detected (high risk: 0.85). The required runtime workflow runs retriever query, which ingests and returns text extracted from the user-provided PDF corpus (outsider-authored document contents) into the agent’s LLM context via the tool output/JSON hits.

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

• Potentially malicious external URL detected (high risk: 0.90). references/install.md instructs cloning and installing from the remote repository https://github.com/NVIDIA/NeMo-Retriever.git (git clone ... then pip install from the checked-out code), which fetches and results in execution of remote code during the install/setup phase if the retriever CLI is missing, so this is a runtime external dependency that can execute code.