Six chapters. Six GH-600 domains. Build, tool, remember, evaluate, coordinate, and govern agents on GitHub-native rails — then prove it in the Grand Capstone.

This campaign is exam prep for GH-600: Developing in Agentic AI Systems. Domain weights below are ranges from Microsoft Learn (“Skills at a glance”) — Learn does not publish fixed percents. Passing score: 700 or greater. Certification renews annually via a free Learn assessment.

Quest objectives

• D1 — Agents in the SDLC — integrate agents in the lifecycle; separate plan vs execute; observability and control
• D2 — Tool Use & Environment — select tools and permissions; configure MCP; safe execution and retries
• D3 — Memory, State & Execution — memory strategies; persist state and detect drift; continuity across tools
• D4 — Evaluation & Tuning — success criteria and signals; failure root-cause analysis; tune from evidence
• D5 — Multi-Agent Coordination — orchestration; multi-agent observability; failure recovery; agent lifecycle
• D6 — Guardrails & Accountability — autonomy levels; guardrails and human-in-the-loop

You will know you are ready when you can map a task to the six domains, write a least-privilege permissions: block and an MCP allow-list, diagnose a failing agent from logs, and keep irreversible actions behind an explicit human gate (the Warden Pact).

Campaign — six chapters

Play in order. Weights cite Learn ranges only.

Source: GH-600 study guide — Skills at a glance.

Every chapter ends in a hands-on lab (mostly bash / jq / gh). The Grand Capstone gates behind all six chapters.

Campaign XP (front matter): hub progression_points: 200; chapters and capstone sum to roughly +600 across the series. The level-banner 6000–7000 range comes from the binary level hub (1100 in _data/quests/levels.yml), not this campaign’s score — treat front-matter progression_points as the source of truth for Agentic Codex XP.

Start here

→ Chapter I — Initiation Rites: Agents in the SDLC

Optional map: GH-600 Study Hub · Skills checklist · Implemented in IT-Journey · Grand Capstone

Study apparatus

• GH-600 Study Guide (Microsoft Learn) — official skills-measured source of truth
• GitHub Copilot coding agent docs
• Model Context Protocol specification
• Extending Copilot with MCP
• GitHub Actions documentation
• GitHub Models

Campaign completion checklist

• Completed all six chapters in order
• Broke every domain seal (D1 through D6)
• Earned the Warden of Autonomy and Codex Master badges
• Passed the Grand Capstone with a governed multi-agent system