Case B
Building the Digital Workforce
Scaling agentic capability via MCP training — the AI KT Series.
01 / Pipeline
- curriculum
- mcp_server
- custom_instructions
- Human checkpoint: hitl_gate
- shipped_tool
02 / Problem
Engineers had AI assistants but no shared patterns for extending them. Every team reinvented prompts, and nobody owned when an agent should stop and ask.
03 / Approach
- 01Designed a curriculum: MCP fundamentals → building tools → Skills and IDE optimization → agentic workflows.
- 02Documented custom instructions that gate which skills an agent may use, per repo and per task.
- 03Standardized Human-in-the-Loop checkpoints so every generated artifact has a named reviewer.
- 04Turned workshop outputs into shipped internal tools, not slideware.
04 / Governance hitl
Custom instructions define agent scope; HITL checkpoints are mandatory before anything leaves the developer's machine.