Everything you need to work through the bootcamp in one place: the course map, masterclass assets, curated backlog examples, and five artifact templates.
Start here, then keep this open all week.
The bootcamp is one continuous engineering story. You follow an engineer joining a project and gradually building a mature AI-assisted delivery workflow:
Task / Work Item
→ Spec
→ Plan
→ Context
→ Team Toolset
→ MCP / Subagents
→ Verification
→ Safe Refactoring
→ Release Readiness
→ Review & Merge Decision
- Read the course map below to see how each session hands off to the next and which artifact it feeds.
- Select a ticket for implementation using the curated backlog description and examples.
- Work through the checklists in
artifacts-todo.md— one per artifact, each ending with a self-check against the evaluation criteria. - Optional: review the masterclass resources in the
mk-resourcesfolder.
Submission: five artifacts by 5 PM on Friday of Week 2, with evidence links (not just prose).
Each masterclass reveals the next engineering problem and adds evidence toward specific Week-2 artifacts.
| # | Masterclass | The problem it solves | Feeds |
|---|---|---|---|
| MK1 | Everything Before Coding Is Context | First project, first task — you discover that everything before coding is context, and that more context isn't automatically better. | (shared mental model) |
| MK2 | Plan Mode & Controlled Implementation | Plan Mode can't invent missing business context. You improve the planning input, review the plan, cut scope creep, and approve only the first safe step. | Artifact 3, 4 |
| MK3 | Context Engineering: Hot / Warm / Cold Layers | A bug in an unfamiliar area — you build a reviewable context map instead of dumping everything in. | Artifact 3, 5 |
| MK4 | Cursor Team Toolset: Rules, AGENTS.md, Skills & Commands | The same guidance is retyped in every chat — you convert stable context into reusable project-level configuration. | Artifact 1 |
| MK5 | MCP & Subagents in Cursor | Manual copying of Jira/GitHub/logs — you choose CLI vs API vs MCP vs Subagent and build one bounded, safe capability. | Artifact 2 |
| MK6 | Verification-First Development with Lightweight Eval Rubrics | "Looks correct" isn't enough — you map acceptance criteria to proof and gate on evidence. | Artifact 3, 4 |
| MK7 | Safe Refactoring, Architecture & Migration Work | A refactor request — you classify the work, lock current behavior, stage it, and approve only the first safe step. | Artifact 3, 4 |
| MK8 | Safe Release for AI-Generated Code | Green checks aren't release readiness — you prepare observability, failure modes, flag decision, and rollback. | Artifact 4 (+ Artifact 5) |
| MK9 | AI-Assisted Code Review & Production-Ready PR | "Tests are green — can we merge?" — you review the evidence package, classify findings, and apply an explicit, human-owned merge decision. | Artifact 3, 4 |
Every session now follows the same rhythm: a realistic project moment → a first attempt that falls short → the framework → a live demo tied to a decision → an explicit human review point → a before/after → the artifact it feeds → a durable takeaway.
Week 2 is your build sprint: turn this experience into the five submitted artifacts.
Full checklists and self-checks are in artifacts-todo.md.
| # | Artifact | What it is | Quality bar |
|---|---|---|---|
| 1 | Specs, Skills & Cursor Project Configuration | Reusable specs, .cursor/rules, AGENTS.md, Cursor Skills (SKILL.md), optional commands |
Reusable by another engineer without extra explanation |
| 2 | Working Agents & MCP Integrations | At least one custom Cursor subagent or MCP integration | Safe by default; read-only preferred unless write is justified |
| 3 | Agentic Workflow End-to-End | Documented flow: spec → plan → code → tests/checks → deployable PR | Shows the full chain from story to production-ready PR |
| 4 | Deployed Deliverable (Ship It) | Working code, deployed or deployable, with verification, release notes and a rollback plan | Merge-ready or demo-ready with evidence it works |
| 5 | Prioritized AI Use Cases & Value Stream Map | 5+ AI use cases with impact/risk/effort ratings, a playbook, and a value stream map (SDLC.json) | Prioritized and actionable, not a generic brainstorm |
