adding-model-support
Pass
Audited by Gen Agent Trust Hub on Apr 19, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill provides numerous example CLI commands for users to execute, including model conversion scripts, logits comparison tests, and generation verification using
uv run python. These reference local files and scripts within the Megatron-Bridge repository and represent legitimate development workflows. - [EXTERNAL_DOWNLOADS]: Instructions include fetching model configuration and weight files from HuggingFace (
huggingface.co), which is a well-known and standard service for machine learning assets. This activity is a core requirement for the skill's purpose of onboarding external models. - [PROMPT_INJECTION]: The content is purely instructional and technical. It does not contain any patterns intended to manipulate agent behavior, override previous instructions, or bypass safety guidelines. Terms like 'IMPORTANT' and 'CRITICAL' are used appropriately to highlight configuration requirements.
- [REMOTE_CODE_EXECUTION]: No remote code execution vulnerabilities were found. The skill guides the user through the creation of local Python classes and configuration files necessary for model integration.
Audit Metadata