axolotl
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONREMOTE_CODE_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides detailed instructions for fetching datasets, model weights, and configurations from established cloud storage providers including AWS S3, Google Cloud Storage, Azure Blob Storage, and Oracle Cloud Infrastructure, as well as the Hugging Face Hub.
- [COMMAND_EXECUTION]: The documentation includes various command-line examples for executing LLM training, running inference loops, and performing hardware performance benchmarks using tools like NCCL Tests and the Axolotl CLI.
- [REMOTE_CODE_EXECUTION]: The documentation references standard installation scripts for well-known development tools, such as the 'uv' package manager via its official domain (astral.sh), which is common in modern Python development environments.
- [INDIRECT_PROMPT_INJECTION]: As the skill is designed to assist in processing external datasets for model training, it operates within an environment exposed to indirect prompt injection. The documentation mitigates this by emphasizing the use of structured chat templates and boundary delimiters (e.g., ChatML) to correctly scope assistant responses.
Audit Metadata