command-creation

SUSPICIOUS. The skill’s purpose is broadly legitimate, but it contains literal pre-execution !`...` directives that would run local commands at skill load/render time, creating an execution risk disproportionate to a documentation-style helper. It also contains misleading guidance about forbidding model despite official docs and its own examples allowing it. No credential theft, remote installer, or exfiltration path is evident, so this is not confirmed malware.

This fragment is a command-template documentation file for an AI-assisted development workflow. It does not itself demonstrate malware or intentional data theft. However, if a runner executes the embedded shell blocks, unvalidated user arguments can enable command injection, and several templates perform destructive, repository-modifying, dependency-installing, or publishing actions without strong confirmation and sandboxing. Treat the file as potentially unsafe automation configuration and require argument escaping, operation allowlisting, explicit approval for destructive or remote actions, and execution in a restricted environment.