The Agent Skills Directory

[SAFE]: The skill is primarily instructional documentation for the LitGPT library. No malicious code, obfuscation, or persistence mechanisms were detected.
[EXTERNAL_DOWNLOADS]: The skill instructs the user to download the litgpt library from PyPI and model weights from well-known and trusted organizations such as Meta, Microsoft, and Google through the Hugging Face model zoo.
[COMMAND_EXECUTION]: The skill documents the use of standard machine learning CLI tools and scripts, including the litgpt command-line interface, torchrun for distributed training, and SLURM for cluster job submission. These are intended behaviors for the skill's primary purpose.
[PROMPT_INJECTION]: The skill facilitates an indirect prompt injection surface by processing user-provided datasets for model fine-tuning and inference.
Ingestion points: External JSON datasets (e.g., data/my_dataset.json) containing instructions and responses.
Boundary markers: Structural separation is maintained through the JSON data format (fields for instruction, input, and output).
Capability inventory: The skill provides instructions for executing model fine-tuning (litgpt finetune) and text generation (litgpt generate) using these datasets.
Sanitization: The documentation does not specify sanitization or filtering of the dataset content, which is typical for model training workflows.

implementing-llms-litgpt