implementing-llms-litgpt
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions for downloading pretrained model architectures and weights using the
litgpt downloadcommand. These downloads target well-known and reputable organizations including Microsoft (phi-2), Meta (Llama), and Google (Gemma). These are standard resources in the machine learning ecosystem and are treated as safe sources. - [COMMAND_EXECUTION]: Instructions include shell commands for installing the
litgptpackage via pip, preparing datasets, and executing fine-tuning or pretraining workflows. These commands use standard tools and the official library's CLI, presenting no malicious intent or unexpected execution patterns. - [INDIRECT_PROMPT_INJECTION]: The skill defines a surface for processing untrusted data by instructing the agent on how to fine-tune models using custom JSON datasets. While this ingestion of external data is an attack vector for indirect prompt injection, it is the primary intended function of the skill for model training. The provided examples follow standard data formatting conventions without bypassing safety controls.
Audit Metadata