paper2video

SUSPICIOUS: the skill’s media-generation capabilities mostly match its stated purpose, but its footprint expands through a third-party personal GitHub dependency and an external skill/workflow trust chain. This is not confirmed malware, yet the install/execution trust model is broader than a self-contained video skill and warrants medium risk.

No clear embedded backdoor or credential-stealing logic is present. However, the code has meaningful security risks consistent with a suspicious rendering/muxing tool: it executes external binaries (LibreOffice/ffmpeg/Chrome) whose paths can be influenced by environment variables/CLI, it renders inlined SVGs with hrefs preserved for http(s)/file/data schemes in a headless browser context (potential SSRF/LFI-like behavior during rendering), and it constructs ffmpeg -filter_complex expressions from untrusted visual-cues JSON (DSL injection risk within ffmpeg parsing/filter behavior). Overall, this module should be treated as security-sensitive and not used with untrusted project assets or untrusted environment configuration.