image-skill

SUSPICIOUS: the skill's core purpose matches image generation, but it routes all work through a proprietary hosted runtime, uses unpinned `npx @latest`, and asks the agent to install a second external skill. The requested token is proportionate, yet the combined hosted mediation, mutable execution path, and transitive trust chain make this higher-risk than a normal documentation-only skill.

SUSPICIOUS: the skill's hosted image-editing purpose generally matches its capabilities, but it carries medium risk because it installs and runs mutable third-party npm code, instructs transitive skill installation from a GitHub slug, and routes user images/prompts to the publisher's hosted API for processing and retention. This looks coherent for a hosted image service, not confirmed malware, but it requires meaningful trust in external code and infrastructure.

SUSPICIOUS. The skill's hosted audio-generation purpose broadly matches its behavior, but it relies on unpinned `npx` execution and transitive skill installation from a third-party repo, with all prompts and outputs routed through a proprietary hosted service. No direct credential theft or clearly malicious behavior is shown, but install trust and third-party data handling make the overall risk medium.