The Agent Skills Directory

[SAFE]: The skill is dedicated to legitimate model analysis tasks. No evidence was found of prompt injection, data exfiltration, or malicious persistence mechanisms.
The instructions strictly guide the agent through technical assessment workflows.
Security boundaries are maintained by focusing only on model metadata and local implementation files.
[COMMAND_EXECUTION]: The skill employs standard system and programming tools for model analysis.
Uses python3 to parse config.json and inspect model implementations.
Employs shell commands such as ls, cat, and pip show to verify the environment and model file integrity.
[EXTERNAL_DOWNLOADS]: Includes instructions for fetching model configuration data from established community platforms.
Recommends using modelscope and huggingface-cli to download non-weight metadata from known repositories.
These operations are standard in AI development and are directed at trusted, well-known services.
[REMOTE_CODE_EXECUTION]: Documentation mentions the trust_remote_code=True parameter within the context of troubleshooting model loading errors.
This is a standard configuration within the Hugging Face ecosystem for supporting custom architectures.
The skill does not automate or force the execution of untrusted remote code without technical justification.

huawei-cloud-msmodelslim-model-analysis