ml-training-recipe
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides high-level instructions for a machine learning research workflow. No security risks were identified.
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to research external sources (papers, datasets, code), which constitutes an ingestion surface for untrusted data. While this is an inherent risk of research tasks, the skill itself does not implement dangerous capabilities that would escalate this risk.
- Ingestion points: External papers, datasets, documentation, and code repositories via the 'researcher' agent.
- Boundary markers: None explicitly defined in the workflow instructions.
- Capability inventory: Mentions writing to an 'outputs/' directory; no subprocess or network capabilities are directly exposed in the skill definition.
- Sanitization: No specific sanitization or validation of the ingested content is described.
Audit Metadata