design-fast

SUSPICIOUS: the core purpose is plausible, but the skill’s footprint relies on transitive skill activation and local home-directory script execution, with at least one referenced dependency (`ai-multimodal`) pointing to a third-party skill install path. Data flow is not overtly exfiltrative, yet the trust chain and external-content-to-file-write pattern make the skill medium-to-high risk for an AI agent environment.