The Agent Skills Directory

[SAFE]: The skill utilizes WebFetch to retrieve technical documentation and model specifications from reputable and trusted domains including huggingface.co, pytorch.org, github.com, and wandb.ai. These network operations are limited to documentation retrieval and are essential for the skill's grounding functionality.
[SAFE]: The skill implements a strict 'Grounding Law' requiring the agent to tag all technical claims with citations, which prevents hallucinations and ensures that advice is derived from verified knowledge bases or official documentation.
[SAFE]: The skill reviews local experiment history via experiments/journal.md, which is a safe and standard practice for context-aware machine learning assistants to provide relevant advice.
[SAFE]: No evidence of prompt injection, hardcoded credentials, malicious obfuscation, or unauthorized remote code execution was found.

ml-iterate