📜 The Agentic Codex

Study Hub for Exam GH-600: Developing in Agentic AI Systems

Deep within the GitHub Citadel, an ancient order guards the Agentic Codex — a tome of forbidden knowledge describing how autonomous agents are summoned, bound, and governed within the Software Development Life Cycle. Only those who master all six disciplines of the Codex earn the title of Agentic Architect.

This is your map.

---

📊 Exam at a Glance

Exam	GH-600: Developing in Agentic AI Systems
Passing Score	700 / 1000
Renewal	Annual (free online assessment)
Stack	GitHub Copilot · GitHub Actions · MCP · GitHub Models API
Official Study Guide	learn.microsoft.com/credentials/certifications/resources/study-guides/gh-600
Learning Paths	Microsoft Learn — Foundations of Agentic AI in GitHub

Domain Weights

#	Domain	Weight	Quest Count
1	Prepare Agent Architecture & SDLC Processes	15–20%	3
2	Implement Tool Use & Environment Interaction	20–25%	4
3	Manage Memory, State & Execution	10–15%	3
4	Perform Evaluation, Error Analysis & Tuning	15–20%	3
5	Orchestrate Multi-Agent Coordination	15–20%	4
6	Implement Guardrails & Accountability	10–15%	3 (inc. capstone)

---

🗺️ Quest Map — The Agentic Codex

graph TD
    Start([🌟 Begin: The Agentic Codex]) --> D1

    subgraph D1["⚙️ Domain 1: Agent Architecture & SDLC (15-20%)"]
        Q1[Q1: Initiation Rites\nSDLC Integration]
        Q2[Q2: The Three Sigils\nPlan vs. Action Boundaries]
        Q3[Q3: The All-Seeing Eye\nObservability & Control]
    end

    subgraph D2["🔧 Domain 2: Tool Use & Environment (20-25%)"]
        Q4[Q4: Forging the Arsenal\nTool Selection & Permissions]
        Q5[Q5: The MCP Conclave\nMCP Server Mastery]
        Q6[Q6: Bind the Realm\nDev Environment Integration]
        Q7[Q7: Shield of Retries\nSafe Execution & Error Handling]
    end

    subgraph D3["🧠 Domain 3: Memory, State & Execution (10-15%)"]
        Q8[Q8: Vaults of Recollection\nMemory Strategies]
        Q9[Q9: Anchoring the Drifting Agent\nState Persistence & Drift]
        Q10[Q10: Crossing Tool Planes\nState Continuity Cross-Tools]
    end

    subgraph D4["📊 Domain 4: Evaluation & Tuning (15-20%)"]
        Q11[Q11: The Oracle's Rubric\nSuccess Criteria & Signals]
        Q12[Q12: The Necromancer's Inquest\nFailure Root Cause Analysis]
        Q13[Q13: Reforging the Agent's Mind\nBehavior Tuning]
    end

    subgraph D5["🕸️ Domain 5: Multi-Agent Coordination (15-20%)"]
        Q14[Q14: The Council of Many\nOrchestration Patterns]
        Q15[Q15: The Scribe's Codex\nMulti-Agent Observability]
        Q16[Q16: When Familiars Fall\nFailure & Recovery]
        Q17[Q17: The Agent Pantheon\nLifecycle Management]
    end

    subgraph D6["🛡️ Domain 6: Guardrails & Accountability (10-15%)"]
        Q18[Q18: The Autonomy Scales\nAutonomy Levels Matrix]
        Q19[Q19: The Warden's Pact\nGuardrails & Human-in-Loop]
        CAP[🏆 CAPSTONE: Trial of the\nAgentic Codex]
    end

    D1 --> D2
    D2 --> D3
    D3 --> D4
    D4 --> D5
    D5 --> D6

    style Start fill:#4CAF50,stroke:#2E7D32,stroke-width:3px,color:#fff
    style CAP fill:#FFD700,stroke:#F57F17,stroke-width:4px,color:#000

---

📚 The Quest Line — Ordered Learning Path

Domain 1 — Prepare Agent Architecture & SDLC Processes

Level: 0111 (Journeyman ⚔️) → builds on API & integration skills

Quest	Title	Difficulty	Time	GH-600 Sub-Skill
Q1	Initiation Rites: Embedding Agents in the SDLC	🟡 Medium	90 min	Integrate agents into the SDLC
Q2	The Three Sigils: Plan, Reason, Act	🟡 Medium	90 min	Define boundaries between planning, reasoning, and action
Q3	The All-Seeing Eye: Observability & Control	🟡 Medium	90 min	Configure observability and control for autonomous agents

Domain 2 — Implement Tool Use & Environment Interaction

Level: 1000 (Expert 🔥) → cloud-native GitHub tooling

Quest	Title	Difficulty	Time	GH-600 Sub-Skill
Q4	Forging the Agent’s Arsenal	🟡 Medium	90 min	Select and configure agent tools
Q5	The MCP Conclave	🔴 Hard	120 min	Configure MCP servers
Q6	Bind the Agent to the Realm	🔴 Hard	120 min	Integrate agents within development environments
Q7	The Shield of Retries	🔴 Hard	120 min	Operate agents with safe execution paths and robust error handling

Domain 3 — Manage Memory, State & Execution

Level: 1001 → 1010 (Expert 🔥) → advanced state management

Quest	Title	Difficulty	Time	GH-600 Sub-Skill
Q8	Vaults of Recollection: Memory Strategies	🔴 Hard	90 min	Implement agent memory strategies
Q9	Anchoring the Drifting Agent	🔴 Hard	90 min	Persist agent state and manage context drift
Q10	Crossing the Tool Planes	🔴 Hard	90 min	Ensure continuity of agent memory and state across tools

Domain 4 — Perform Evaluation, Error Analysis & Tuning

Level: 1010 (Expert 🔥) → evaluation harnesses and signal analysis

Quest	Title	Difficulty	Time	GH-600 Sub-Skill
Q11	The Oracle’s Rubric	🟡 Medium	90 min	Define success criteria and evaluation signals
Q12	The Necromancer’s Inquest	🔴 Hard	90 min	Analyze agent failures and identify root causes
Q13	Reforging the Agent’s Mind	🔴 Hard	120 min	Tune agent behavior based on evaluation results

Domain 5 — Orchestrate Multi-Agent Coordination

Level: 1011 (Expert 🔥) → orchestration and resilience patterns

Quest	Title	Difficulty	Time	GH-600 Sub-Skill
Q14	The Council of Many: Orchestration Patterns	🔴 Hard	120 min	Operate and manage multi-agent workflows
Q15	The Scribe’s Codex: Multi-Agent Observability	🔴 Hard	90 min	Configure observability for multi-agent behavior
Q16	When Familiars Fall: Failure & Recovery	🔴 Hard	90 min	Detect and respond to multi-agent failures
Q17	The Agent Pantheon: Lifecycle Management	⚔️ Epic	120 min	Manage the lifecycle of agents within multi-agent workflows

Domain 6 — Implement Guardrails & Accountability

Level: 1100 (Expert 🔥 → Legend 🏆) → responsible autonomy

Quest	Title	Difficulty	Time	GH-600 Sub-Skill
Q18	The Autonomy Scales	🔴 Hard	90 min	Define autonomy levels
Q19	The Warden’s Pact	🔴 Hard	120 min	Implement guardrails and human-in-the-loop workflows
🏆	CAPSTONE: Trial of the Agentic Codex	⚔️ Epic	240 min	All 6 domains integrated

---

📖 Chronicle Posts

Development sessions documented in real time — see every quest built live.

Post	Domain(s)	Category
Launching the Agentic Codex: A GH-600 Learning Track for GitHub Native AI	All	ai-agents
Embedding Agents in the SDLC — What Actually Changes	D1	ai-agents
MCP Servers and Agent Tooling in Practice	D2	ai-agents
Taming Agent Memory and Context Drift	D3	ai-agents
Evaluating and Tuning Agents with GitHub Signals	D4	ai-agents
Orchestrating Multi-Agent Workflows on GitHub	D5	ai-agents
Agent Guardrails and Responsible Autonomy	D6	ai-agents

---

🗒️ Quick-Reference Notes

Note	Purpose
Exam Overview & Logistics	Scoring, scheduling, renewal
Skills Checklist (Printable)	All 19 sub-skills as checkboxes
Copilot Coding Agent Cheatsheet	Invocation, scopes, autonomous-PR pattern
MCP Quick Reference	Server config, registries, allow-list syntax
Autonomy Levels Matrix	Risk classification decision table
Evaluation Signals Table	Qualitative vs quantitative signals
Glossary	Agent, MCP, HITL, drift, autonomy, …

---

🧰 Recommended Resources

See the full Recommended Resources page for Microsoft Learn paths, GitHub documentation, MCP specification links, and the GitHub Models API reference.

See the Week-by-Week Learning Path for a structured study schedule with time estimates and daily objectives.

---

🕸️ Knowledge Graph

Structured wiki-links enable the Obsidian-style local knowledge graph for this hub. Click the graph button (bottom-left) to explore connections.

Domain 1 — SDLC Integration: [[Initiation Rites: Embedding Agents in the SDLC]] · [[The Three Sigils: Plan, Reason, Act]]

Domain 2 — Tools & Observability: [[The MCP Conclave: Mastering Model Context Protocol Servers]] · [[The All-Seeing Eye: Observability & Control for Autonomous Agents]] · [[Forging the Agent’s Arsenal: Tool Selection & Permissions]]

Domain 3 — Memory & Context: [[Bind the Agent to the Realm: Dev Environment Integration]] · [[Vaults of Recollection: Agent Memory Strategies]] · [[The Shield of Retries: Safe Execution and Error Handling]]

Domain 4 — Evaluation: [[The Oracle’s Rubric: Defining Agent Success Criteria and Signals]] · [[The Necromancer’s Inquest: Agent Failure Root Cause Analysis]] · [[Reforging the Agent’s Mind: Behavior Tuning Through Instructions]] · [[Anchoring the Drifting Agent: State Persistence and Drift Prevention]] · [[Crossing the Tool Planes: State Continuity Across Tools]]

Domain 5 — Multi-Agent: [[The Council of Many: Multi-Agent Orchestration Patterns]] · [[The Scribe’s Codex: Observability in Multi-Agent Systems]] · [[When Familiars Fall: Multi-Agent Failure Recovery]] · [[The Agent Pantheon: Multi-Agent Lifecycle Management]]

Domain 6 — Guardrails: [[The Autonomy Scales: Mapping Agent Autonomy Levels]] · [[The Warden’s Pact: Guardrails and Human-in-the-Loop Patterns]] · [[Trial of the Agentic Codex: The Grand Capstone]]

Reference Notes: [[GH-600 Exam Overview]] · [[GH-600 Skills Checklist]] · [[GitHub Copilot Coding Agent Cheatsheet]] · [[MCP Quick Reference]] · [[Autonomy Levels Matrix]] · [[Evaluation Signals Table]] · [[GH-600 Glossary]] · [[GH-600 Agentic AI Quick-Reference Notes]]

---

🚀 Getting Started

New to this track? Follow these four steps in order:

1. Read the Exam Overview (5 min) to understand format and scoring.
2. Self-assess with the Skills Checklist — tick off what you already know.
3. Pick your pace from the Week-by-Week Learning Path (default: 6 weeks at ~7 hrs/week).
4. Begin with Q1: Initiation Rites — the first quest of Domain 1.

Already comfortable with Copilot agents? Jump to Domain 4 — Evaluation & Tuning for the deepest material.

Cramming for an upcoming exam? Read the Skills Measured breakdown, then attempt the Capstone Trial as a diagnostic.

---

❓ Frequently Asked Questions

Do I need a paid GitHub Copilot subscription?
Most quests work with the free Copilot trial. A small number of multi-agent and coding-agent quests benefit from an active Copilot Business or Enterprise plan but are not blocked without one.

How long does the full track take?
About 40 hours of focused study spread over 6 weeks. Faster if you skip the chronicle posts; slower if you build out the optional sandbox in work/gh-600/.

What if the official exam updates?
We track the official GH-600 Study Guide and refresh quests when sub-skills change. The PR that introduced this track (#272) is the baseline; subsequent updates live in the repository’s release notes.

Is the capstone equivalent to the real exam?
No — it is a diagnostic, not a practice exam. It surfaces blind spots; it does not predict your score.

Can I contribute fixes or additional quests?
Yes. See the contribution guidance on the certifications hub and open an issue first to align scope.

---

📜 Track Maintenance

Last reviewed	2026-05-17
Validator status	21/21 quests passing
Mermaid rendering	Enabled on all 20 quest pages
Wiki-link graph	28 edges from hub, 2–3 per note
Issue tracker label	cert:gh-600

Spotted a stale link or outdated reference? Open an issue with the cert:gh-600 label so it joins the next maintenance pass.