🛡️Understand Prompt Injection Attacks
Audit your own LLM features for injection surfaces. Separate direct from indirect attacks with worked examples, then apply structured isolation, output filters, provenance, and least-authority tool design.
Phase 1The Trust Boundary Your Prompt Quietly Erases
See trust boundaries that prompts silently erase
Your system prompt isn't a wall — it's a suggestion
6 minYour system prompt isn't a wall — it's a suggestion
Direct injection is a user typing past your guardrails
6 minDirect injection is a user typing past your guardrails
Indirect injection: the attacker isn't even your user
7 minIndirect injection: the attacker isn't even your user
There is no parser that separates instructions from data
6 minThere is no parser that separates instructions from data
Phase 2Reproduce the Two Attack Shapes
Reproduce direct and indirect injections on a toy app
Build a 50-line target before you can reason about defense
7 minBuild a 50-line target before you can reason about defense
Land a direct injection on your own toy app
8 minLand a direct injection on your own toy app
Plant a payload in a webpage and let your app find it
8 minPlant a payload in a webpage and let your app find it
Five ways a successful injection can hurt you
7 minFive ways a successful injection can hurt you
Write a one-page threat model for your toy app
7 minWrite a one-page threat model for your toy app
Phase 3Defenses That Actually Hold
Apply isolation, filters, provenance, and least authority
When a scenario calls for structured isolation
8 minWhen a scenario calls for structured isolation
An output filter caught what the model didn't
8 minAn output filter caught what the model didn't
Provenance is the missing label on every prompt token
8 minProvenance is the missing label on every prompt token
Your tool list is your real attack surface
8 minYour tool list is your real attack surface
Phase 4Audit Your Own LLM Feature
Audit one of your real LLM features end to end
Audit one real LLM feature end to end
8 minAudit one real LLM feature end to end
Frequently asked questions
- What is prompt injection in LLM applications?
- This is covered in the “Understand Prompt Injection Attacks” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- What's the difference between direct and indirect prompt injection?
- This is covered in the “Understand Prompt Injection Attacks” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- Why can't I just tell the model to ignore injected instructions?
- This is covered in the “Understand Prompt Injection Attacks” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- How is prompt injection different from jailbreaking?
- This is covered in the “Understand Prompt Injection Attacks” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
- What defenses actually work against indirect prompt injection?
- This is covered in the “Understand Prompt Injection Attacks” learning path. Start with daily 5-minute micro-lessons that build from fundamentals to hands-on application.
Related paths
🐍Python Decorators Introduction
Build one mental model for Python decorators that covers closures, argument passing, functools.wraps, and stacking — then ship a working caching or logging decorator from scratch in under 30 lines.
🦀Rust Lifetimes Explained
Stop reading `'a` as line noise and start reading it as scope arithmetic — one failing snippet at a time — until you can thread lifetimes through a small parser or iterator adapter without fighting the borrow checker.
☸️Kubernetes Core Concepts
Stop drowning in 30+ resource types. Build the mental model one primitive at a time -- pods, deployments, services, ingress, config -- then deploy a real app with rolling updates and health checks.
📈Big O Intuition
Stop treating Big O as math you memorized for an interview — build the intuition to spot O(n²) disasters, pick the right data structure without thinking, and rewrite a slow function from O(n²) to O(n) in under five minutes.