Reusable AI Workflows with Agent Skills скачать в хорошем качестве

Reusable AI Workflows with Agent Skills

In this video, we'll walk through a reusable metadata extraction workflow built as an Agent Skill — no custom application code required. Instead of re-prompting or rebuilding the same logic over and over, this approach packages a real capability (“extract metadata from documents in Box and write it back safely”) into a declarative skill that can be installed once and reused anywhere. The demo shows how: ✅ An Agent Skill defines behavior in a structured SKILL.md ✅ The agent uses Box MCP to read files directly from Box ✅ Box AI extracts structured values based on a metadata template ✅ Metadata is written back safely without overwriting existing values ✅ Files remain in Box as the system of record the entire time This isn’t about generating code — it’s about orchestration. The agent discovers the right context, invokes the right tools, and performs real work on real files. If you’re interested in building reusable, tool-driven AI workflows — especially ones grounded in schemas, permissions, and enterprise content — this pattern is worth a look. Open-source Agent Skill (SKILL.md): https://github.com/box-community/box-... Topics covered: ✅ Agent Skills ✅ Box MCP ✅ Box AI ✅ Metadata-driven workflows ✅ Declarative agent behavior ✅ Reusable AI orchestration