The Agent Skills Directory

[SAFE]: The skill provides instructions and patterns for using the react-native-executorch library, which is a legitimate open-source project maintained by Software Mansion. All referenced packages and repositories are consistent with the vendor's identity.
[EXTERNAL_DOWNLOADS]: The documentation describes fetching AI models from huggingface.co/software-mansion. HuggingFace is an established service for AI model hosting, and the repository belongs to the vendor. These downloads are standard for on-device AI applications and do not involve untrusted sources.
[COMMAND_EXECUTION]: The skill includes standard installation commands using npm install. There are no instances of suspicious or unauthorized shell command execution, such as privilege escalation or persistence mechanisms.
[DATA_EXFILTRATION]: No patterns of sensitive data access or exfiltration to external domains were identified. The library focuses on on-device processing, which is inherently privacy-preserving as described in the metadata.
[REMOTE_CODE_EXECUTION]: No evidence of dynamic code execution from untrusted remote sources or patterns such as piping curl output to a shell were found.

on-device-ai