nvidia-speech-nim

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill facilitates the processing of external data which represents a surface for indirect prompt injection.\n
  • Ingestion points: Processes audio files for transcription (ASR) and text for translation (NMT) or synthesis (TTS) as outlined in SKILL.md.\n
  • Boundary markers: Recommends using <dnt> tags to protect specific terminology from modification during translation.\n
  • Capability inventory: Employs curl for HTTP requests and the nvidia-riva-client library for gRPC interactions with microservices.\n
  • Sanitization: Uses terminology control and tags to manage content, though it lacks explicit instructions for filtering instructions embedded within audio transcripts or input text.\n- [EXTERNAL_DOWNLOADS]: The skill utilizes the official NVIDIA Riva client for Python.\n
  • Evidence: Recommends pip install -U nvidia-riva-client for interacting with Speech NIM services.\n
  • Context: The package is provided by a well-known technology vendor and hosted on standard package registries.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 06:42 PM