huashu-design

SUSPICIOUS. The core capabilities mostly fit a design-production skill, and the install paths are largely official registry/open-source tools rather than obvious malware delivery. However, the skill has a wide execution footprint, uses many unpinned external tools, and includes optional cloud scripts that may forward user media and API keys to third-party services without clear endpoint disclosure, making it higher risk than a normal documentation-style skill.

No evidence of overt malware/backdoor behavior is present in this module. However, the code intentionally performs authenticated off-host transmission of user-provided video content (base64 data URLs) and structured prompt/context (including optional director context and derived timeline metadata) to a hardcoded third-party AI API, then persists the remote output locally in a markdown report. This creates a substantial privacy and supply-chain operational risk if users/teams do not explicitly consent to external disclosure and data retention. Additionally, the snippet shows CHECKLIST/SEVERITY_RULE placeholders, which could indicate the provided fragment may be incomplete and reduces confidence in exact runtime behavior.