nemotron-asr-finetune
Pass
Audited by Gen Agent Trust Hub on Sep 16, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill workflow involves downloading resources from the official NVIDIA GitHub organization (e.g.,
github.com/NVIDIA-NeMo/Speech.git) to set up local training environments. These operations are transparent, targeted at official vendor sources, and consistent with the intended functionality for machine learning model development. - [COMMAND_EXECUTION]: The skill includes a shell script (
link-nemo-subskill.sh) designed to manage the discovery of project-local skills. It performs standard file system operations, such as creating symbolic links within the agent's local skill directory (~/.claude/skills/), which is required to integrate repository-specific tools into the agent's environment. - [INDIRECT_PROMPT_INJECTION]: As an orchestrator, the skill ingests external data including audio manifests and training transcripts. The workflow incorporates several security best practices, such as pre-flight quality checks, structured reporting via an experiment ledger, and explicit user confirmation gates before proceeding to high-resource training stages, which effectively reduces the risk of unintended behavior from malicious data inputs.
Audit Metadata