Study Group Kit

June 16, 2026 ยท View on GitHub

Use this kit to run the Agent Engineering Roadmap as a study group, internal training, or community workshop.

Formats

FormatBest forCommitment
4-week sprintExperienced engineers2 sessions per week
8-week cohortMixed skill levels1 session per week
1-day workshopTeam kickoff4 focused blocks

4-Week Sprint

WeekFocusRunDeliverable
1Foundations, tools, MCPExamples 01-03Agent spec and tool contract
2Memory, RAG, workflowExamples 04, 05, 08RAG eval and workflow sketch
3Graphs, multi-agent, approvalExamples 06, 09Approval graph and routing tests
4Production, observability, securityExamples 10, 11, benchmarkTrace review and injection defense

8-Week Cohort

WeekTopicDiscussion Prompt
1What is an agent?Where does a chatbot stop being enough?
2Tool use and MCPWhich tools should never be callable without approval?
3MemoryWhat should an agent forget?
4RAGHow do we know an answer is grounded?
5Workflow graphsWhich states should be explicit?
6Multi-agent systemsWhen does adding agents make quality worse?
7ObservabilityWhich trace would explain a bad answer?
8Security and capstoneWhat attack would break your design?

Session Template

  1. Open with one failure story.
  2. Draw the mental model.
  3. Run the smallest example.
  4. Change one thing and predict the result.
  5. Run the benchmark or eval.
  6. Record one improvement as an issue.

Facilitator Checklist

  • Pick one runnable example before each session
  • Prepare one failure case
  • Keep discussion tied to observable behavior
  • Ask learners to submit one artifact each week
  • End each session with one concrete repo improvement