The Agent Skills Directory

[SAFE]: No malicious patterns detected. The skill is designed to provide a structured verification process for machine learning configurations and code snippets using documented framework behaviors.
[EXTERNAL_DOWNLOADS]: The skill fetches documentation from well-known and trusted sources, including Hugging Face, PyTorch, DeepSpeed, and vLLM. These external references are used to cross-check user-provided parameters against official API signatures and recommended ranges, which is consistent with the skill's stated purpose.
[PROMPT_INJECTION]: The skill employs strict instructional constraints and 'Iron Law' directives to ensure high-fidelity output and grounding. These constraints are internal safety measures to prevent hallucinations and maintain citation integrity, rather than attempts to bypass agent safety filters or override core behavioral guidelines.

ml-verify