The Agent Skills Directory

[SAFE]: No malicious code, persistence mechanisms, or unauthorized privilege escalations were detected within the skill documentation or code examples.\n- [EXTERNAL_DOWNLOADS]: The instructions reference the nvidia-riva-client library and NVIDIA's official GitHub repositories. These are legitimate resources provided by a well-known technology service.\n- [COMMAND_EXECUTION]: Technical examples use curl to interact with services hosted on localhost, representing standard local development and testing workflows.\n- [PROMPT_INJECTION]: The skill identifies the attack surface for indirect prompt injection when processing untrusted audio and text. It mandates several mitigation strategies: ingestion through formal APIs (ASR/TTS NIM), use of boundary markers like <dnt> tags for translation, a capability inventory involving the Riva client, and sanitization through audio normalization and terminology dictionaries.

nvidia-speech-nim