The Agent Skills Directory

[SAFE]: The skill implements a comprehensive safety boundary for browser automation. Specifically, the armFinalPublishGuard function in scripts/v2/ego/core.mjs injects a capture-phase guard to prevent accidental or unauthorized clicks on final 'Publish' buttons across all supported platforms.
[SAFE]: User configuration and recurring preferences are handled securely. As documented in references/configuration.md and implemented in scripts/lib/config.mjs, the skill stores per-user settings in restricted directories (e.g., $XDG_CONFIG_HOME) using 0o600 file permissions and explicitly avoids persisting sensitive data like cookies, credentials, or session tokens.
[SAFE]: The architecture utilizes multiple levels of atomic locking (publisher, job, and platform levels) and one-generation state backups to prevent race conditions and data corruption during parallel operations.
[SAFE]: The skill demonstrates a high level of operational integrity by performing pre-flight media validation (e.g., checking video duration via ISO BMFF metadata in scripts/lib/media.mjs) before starting browser sessions, ensuring that invalid assets are rejected early without wasting resources.
[SAFE]: All external resource interactions are confined to legitimate creator platform domains (Xiaohongshu, Douyin, Bilibili, and WeChat Channels) and their respective content delivery networks. No suspicious data exfiltration or unauthorized network patterns were identified.

video-publisher