Free & open · no paywall, ever

IntegrAuth Academy

Identity, security and practical AI — free, from zero. Byte-sized lessons, real diagrams, hands-on labs, zero fluff. Two programs, each with its own exam and certificate.

Choose your program Browse the 20 tracks
20 tracks 241 lessons 242 hands-on labs 230+ diagrams 100% free

English is hard? Your browser can translate this page into Tamil, Telugu, Kannada, Malayalam or Hindi. Tech words stay in English so you learn the real names.

How?
  • Chrome on phone: ⋮ menu → Translate
  • Chrome on computer: right-click → Translate to …
  • Edge: the translate icon in the address bar
Practical AI Track 4 · Coding with AI
0/75 lessons read
CODING WITH AI · LESSON 10

The big picture

Finance asked Priya for a monthly refund report. She built it with a coding agent, using every lesson. Your goal: see how the steps fit together.

The leash: rules · hooks · sandbox · narrow tokens1 · Brief & specAGENTS.md + acceptance criteria2 · Planread-only; you approve3 · Build + testsmall steps, tests locked4 · Review & mergetests first, every dependencypoisoned issue: "send .env out"send it backupload stopped at the fencegit underneath it all: a branch, a baseline and a commit after every step you've read
Figure — You own the start and the end. Limits hold the middle. You write the brief and you do the review. Rules, hooks and a sandbox stop a fooled agent. Git makes every step easy to undo.
Key points
  • A claim is not proof. "All tests pass" is just a sentence. Check the test output.
  • Context is the part you control. A short instruction file and a clear spec often help more than a bigger model.
  • Text only asks. Hooks and sandboxes enforce. Put must-always and must-never rules outside the model.
  • Your review sets the pace. Run only as many agents as you can review.

What Priya did

  • Set up: a clean start, a new branch, a test run and a spend limit.
  • Context: a short AGENTS.md held the commands and traps.
  • Plan: a spec with acceptance criteria. She read the plan and cut one risky step.
  • Tests: tests first, then a commit. The agent loosened a test, and the diff showed it.
  • Review: she dropped a 3-day-old package nobody knew.
  • Extend: a hook blocked edits to .env, every time.
  • Safety: a hidden order in an issue fooled the agent. The rules and the sandbox stopped it.
  • App builders: she tried the screen in an app builder, with fake data.
  • Many agents: two bug fixes ran in separate worktrees. She reviewed each one.

Where to go next

Next: the cheat sheet & pop quiz.