DEV-104Beginner to expert
Developer Path · Step 4 of 8
Foundations, Re-Engineered
Relearn the fundamentals at senior depth: the agentic loop, prompts-as-specs, and secrets hygiene from a cybersecurity-first team.
What you’ll walk away with
- Explain and control the agentic loop — how tools, context, and permissions shape what an AI agent actually does
- Set up a project for power using CLAUDE.md and tight permission scopes that create a fast, trustworthy feedback loop
- Write prompts as specs — deterministic, constrained, and evaluable rather than vague natural-language requests
- Script real automations with idempotency, CLIs, and Node tooling so runs are safe to repeat and easy to trust
- Debug AI-generated code using stack traces, logs, and reproductions instead of guesswork
- Wire Claude into your own stack via MCP and tool integration, and apply least-privilege secrets and supply-chain hygiene to keep it safe
Where this leads
AIU Prism · Production Architect (AIU-PA)
You can take a feature from spec to production with Claude Code — architected, secured, tested, and shipped.
Questions people ask before they start
Is this course for total beginners or people who already know some AI coding basics?
It's built for people who've touched AI coding tools but want the fundamentals taught at senior depth — things like the agentic loop, context control, and prompts-as-specs go well beyond a first introduction.
Why does a cybersecurity company teach an AI coding course?
Securafy is a security company first, so this course treats secrets hygiene, least-privilege access, and supply-chain risk as core curriculum, not an afterthought — it's a dedicated module, not a footnote.
Will I actually build things, or is this just theory?
You'll do real project setup with CLAUDE.md and permissions, write scripted automations with CLIs and Node tooling, and debug real code using stack traces and logs — the lessons are hands-on, not conceptual overviews.
Does this cover connecting AI to my own tools and systems?
Yes — there's a dedicated lesson on MCP and tool integration that walks through wiring Claude into your own stack.
What does 'prompts as specs' actually mean?
It's a lesson on treating prompts like engineering specifications — focusing on determinism, constraints, and evals so outputs are predictable and testable instead of one-off guesses.
What's the last thing I'll learn, and where does it leave me?
The course ends with 'From Scripts to Systems: An Engineering Growth Path,' which frames how to grow from writing one-off scripts toward building larger, more reliable systems with AI.
Developer Path track
This course is step 4 of 8. Finish all 8 to earn AIU Prism · Production Architect.
- Step 1Claude Chat, Desktop, Code, and CoWork: What's the DifferenceBLD-101 · 9 lessons
- Step 2Claude Code as a Production ToolDEV-102 · 13 lessons
- Step 3Plan, Build & Ship with Claude CodeDEV-103 · 13 lessons
- Step 4Foundations, Re-EngineeredDEV-104 · 16 lessons
- Step 5The Workshop, Production EditionDEV-105 · 16 lessons
- Step 6MVP to Production (Developer Edition)DEV-106 · 15 lessons
- Step 7Engineered RecipesDEV-107 · 14 lessons
- Step 8Version Control & CI/CD MasteryDEV-108 · 13 lessons
- EarnProduction ArchitectPrism · Architect
The full track
Included with AI University Pro
Course path · DEV-104 · Foundations, Re-Engineered
Adv · Foundations, Re-Engineered
- DEV-104-1.1PreviewLessonThe Agentic Loop Under the Hood: Tools, Context & Control
- DEV-104-1.2LockedLessonProject Setup for Power: CLAUDE.md, Permissions & a Tight Feedback Loop
- DEV-104-1.3LockedLessonPrompts as Specs: Determinism, Constraints & Evals
- DEV-104-1.4LockedLessonScripting Real Automations: Idempotency, CLIs & Node Tooling
- DEV-104-1.5LockedLessonDebugging with Stack Traces, Logs & Reproductions
- DEV-104-1.6LockedLessonMCP & Tool Integration: Wiring Claude Into Your Stack
- DEV-104-1.7LockedLessonSecrets, Least-Privilege & Supply-Chain Hygiene
- DEV-104-1.8LockedLessonFrom Scripts to Systems: An Engineering Growth Path
- DEV-104-1.9LockedPrompt LabHand Claude a Verifiable Task, Not a Wish
- DEV-104-1.10LockedHands-On LabFix a Fictional Booking API With a Scoped Feedback Loop
Intermediate · Context, Sessions & Parallel Work
- DEV-104-2.1LockedLessonManaging Context and Sessions: /context, /compact, /clear, Resume, and Worktrees
- DEV-104-2.2LockedPrompt LabWrite a Hand-Off Note That Survives a Fresh Session
- DEV-104-2.3LockedHands-On LabRun Two Fictional Features in Parallel With Worktrees
Expert · Sandboxing & Safe Autonomy
- DEV-104-3.1LockedLessonSandboxing, Auto Mode, and Deny-by-Default: Guardrails That Hold
- DEV-104-3.2LockedPrompt LabThreat-Model Your Claude Code Setup Before You Automate It
- DEV-104-3.3LockedHands-On LabLock Down a Fictional Repo for an Unattended Overnight Run
What you earn at the end of this track
AIU Prism · Production Architect
A verifiable AI University credential, Prism Architect level. Not a vendor certification.

