{"skill":{"slug":"ezviz-open-multimodal-analysis","displayName":"Ezviz Open Multimodal Analysis","summary":"萤石多模态理解技能。通过设备抓图 + 智能体分析接口，实现对摄像头画面的 AI 理解分析。 Use when: 需要对监控画面进行智能分析、场景识别、行为理解、物体检测等多模态 AI 分析任务。 ⚠️ 安全要求：必须设置 EZVIZ_APP_KEY 和 EZVIZ_APP_SECRET 环境变量，使用最小权限凭证。","tags":{"latest":"1.0.1"},"stats":{"comments":0,"downloads":412,"installsAllTime":0,"installsCurrent":0,"stars":1,"versions":2},"createdAt":1773395727775,"updatedAt":1777529316237},"latestVersion":{"version":"1.0.1","createdAt":1773839994256,"changelog":"**Summary: This update introduces token caching and enhanced credential/fallback handling for improved security and efficiency.**\n\n- Added `lib/token_manager.py` for centralized token management.\n- Implements secure access token caching on disk (`/tmp/ezviz_global_token_cache/global_token_cache.json`), configurable via `EZVIZ_TOKEN_CACHE`.\n- The skill now supports credential fallback: uses environment variables first, then OpenClaw config files as fallback.\n- SKILL.md updated: clarifies token cache, channel config, security recommendations, credential sourcing, and file permissions.\n- No breaking changes; prior environment variable usage remains compatible.","license":"MIT-0"},"metadata":{"os":null,"systems":null},"owner":{"handle":"ezviz-open","userId":"s174s8a00r18xwt2q5p7j85f5183gavv","displayName":"EzvizOpenTeam","image":"https://avatars.githubusercontent.com/u/16202450?v=4"},"moderation":{"isSuspicious":true,"isMalwareBlocked":false,"verdict":"suspicious","reasonCodes":["suspicious.exposed_secret_literal","suspicious.vt_suspicious"],"summary":"Detected: suspicious.exposed_secret_literal, suspicious.vt_suspicious","engineVersion":"v2.4.5","updatedAt":1777529316237}}