The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (SAFE): The skill contains URLs to arXiv papers and official Hugging Face documentation. It references datasets hosted on the Hugging Face Hub, which is standard practice for ML development.
[COMMAND_EXECUTION] (SAFE): Bash examples provided (e.g., python -m trl.scripts.ppo, accelerate launch) are standard CLI entry points for the TRL and Accelerate libraries.
[REMOTE_CODE_EXECUTION] (SAFE): No instances of piped remote execution or dynamic code evaluation from untrusted sources were found.
[DATA_EXFILTRATION] (SAFE): No hardcoded credentials, sensitive file path access, or unauthorized network calls were detected. The scripts focus on local model training and evaluation.
[PROMPT_INJECTION] (SAFE): The text body is strictly instructional and contains no attempts to bypass safety filters or override system instructions.

fine-tuning-with-trl