mcore-migrate-gpt-to-hybrid
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill provides legitimate technical documentation and bash snippets for migrating machine learning model configurations. The instructions include defensive coding practices, such as warnings about bash globbing hazards when expanding arrays. The content is consistent with the stated purpose of migrating Megatron Core checkpoints.
- [INDIRECT_PROMPT_INJECTION]: The skill describes a workflow that ingests external artifacts like logs, checkpoint metadata, and failure outputs. This establishes a surface for indirect prompt injection if an attacker can manipulate these files. Mandatory Evidence: 1. Ingestion points: Workflow Step 1 (artifact ingestion); 2. Boundary markers: Absent; 3. Capability inventory: Shell command execution for script modification and verification; 4. Sanitization: Absent. The risk is considered safe due to the technical nature of the workflow and the focus on internal model migration.
Audit Metadata