監査履歴

Static analysis flagged 43 external command patterns and 19 URLs, but all are FALSE POSITIVES. The external commands are bash code blocks in documentation showing usage examples, not actual shell execution. URLs are documentation links to inference.sh resources. The skill uses allowed-tools: Bash(infsh *) which properly restricts bash commands to the inference.sh CLI only. No malicious intent detected.

This skill provides a wrapper for the inference.sh CLI to generate AI videos. All detected patterns (external_commands, network URLs, cryptographic references) are false positives - the skill contains only documentation examples and legitimate API endpoint references. No malicious behavior detected. The pipe-to-shell install pattern is standard for CLI tools with documented manual alternatives.