Robot Explains Gallery

OWASP Agentic Top 10

A kid-friendly visual interpretation of the OWASP Top 10 for Agentic Applications — when AI can plan, act, use tools, and cause real consequences.

The OWASP Top 10 for Agentic Applications covers the risks that appear when AI can act — using tools, identity, memory, and permissions, and even working with other agents. This set reimagines those ten agentic risks as friendly visual cards, a companion to the LLM Top 10. For awareness and discussion, not a replacement for the official OWASP guidance.

Difficulty: IntermediatePosters: 10Audience: Intermediate

Two lists, paired — because neither covers the other's ground:

  • Agentic Top 10: risks when AI can act — with tools, memory across sessions, and real consequences.
  • LLM Top 10: risks when AI is part of an app — what it reads, reveals, generates, or mishandles. See the LLM Top 10 →

How do the two lists relate? See the crosswalk →

LLM risk

Robot reads or writes something risky.

Agentic risk

Robot uses tools, identity, memory, or permissions to do something risky.

⇩ Download all posters (ZIP)

Advanced visual cards for older learners & grown-ups

Unlike the rest of the library, these are poster cards rather than full lessons, a friendly visual interpretation of the OWASP Top 10 for Agentic Applications, for awareness and discussion. They're best for older learners and grown-ups, and are not a replacement for the official OWASP guidance.

New here? Start with: What is OWASP? 🎮 Play the Agentic Defender Game LLM vs Agentic →

Infographic: Agent Goal Hijack. Someone twists the robot's goal so it works toward the wrong thing.
#1

Agent Goal Hijack

Someone twists the robot's goal so it works toward the wrong thing.

View poster →
Infographic: Tool Misuse & Exploitation. The robot's tools get used in ways they shouldn't be.
#2

Tool Misuse & Exploitation

The robot's tools get used in ways they shouldn't be.

View poster →
Infographic: Identity & Privilege Abuse. The robot uses the wrong identity or more access than it should.
#3

Identity & Privilege Abuse

The robot uses the wrong identity or more access than it should.

View poster →
Infographic: Agentic Supply Chain. An agent, tool, model, or connection the robot trusts isn't safe.
#4

Agentic Supply Chain

An agent, tool, model, or connection the robot trusts isn't safe.

View poster →
Infographic: Unexpected Code Execution. The robot runs code or commands it was never meant to run.
#5

Unexpected Code Execution

The robot runs code or commands it was never meant to run.

View poster →
Infographic: Memory & Context Poisoning. Bad info sneaks into the robot's memory and shapes what it does next.
#6

Memory & Context Poisoning

Bad info sneaks into the robot's memory and shapes what it does next.

View poster →
Infographic: Insecure Inter-Agent Communication. Robots talking to each other can be tricked, copied, or overheard.
#7

Insecure Inter-Agent Communication

Robots talking to each other can be tricked, copied, or overheard.

View poster →
Infographic: Cascading Failures. One small mistake sets off a chain of bigger ones.
#8

Cascading Failures

One small mistake sets off a chain of bigger ones.

View poster →
Infographic: Human-Agent Trust Exploitation. People trust the robot too much, or the robot is used to trick people.
#9

Human-Agent Trust Exploitation

People trust the robot too much, or the robot is used to trick people.

View poster →
Infographic: Rogue Agents. An agent goes off on its own and stops following the rules.
#10

Rogue Agents

An agent goes off on its own and stops following the rules.

View poster →

About this set & license

Robot Explains is an independent educational adaptation and is not official OWASP guidance. It is based on the OWASP Top 10 for Agentic Applications by the OWASP Agentic Security Initiative (ASI).

The OWASP material is licensed under Creative Commons Attribution-ShareAlike 4.0 (CC BY-SA 4.0), and these adapted posters are shared under the same license.

LLM vs Agentic: how they relate → OWASP LLM Top 10