# AI Agent Delegation Security Challenge

Source: https://tenuo.ai/lab/

> Give each agent only the authority its part of the trip needs. A free, hands-on lab in AI agent delegation security.

This is an interactive, hands-on lab. Each stage page has its own instructions and checks; this file lists the stages.

## Stages

- [Stage 1: One key for everyone](https://tenuo.ai/lab/stage-1): See what a rogue agent can do when every agent shares one credential.
- [Stage 2: Every agent gets its own account](https://tenuo.ai/lab/stage-2): Give each agent its own credential and see which damage that removes and which damage remains.
- [Stage 3: Rules that fit the job](https://tenuo.ai/lab/stage-3): Write permissions narrow enough that every rogue action is blocked and the trip still books.
- [Stage 4: Two travelers, then a handoff](https://tenuo.ai/lab/stage-4): First isolate Alice from Bob; then observe why passing a whole credential gives the next agent too much.
- [Stage 5: Access that travels with the work](https://tenuo.ai/lab/stage-5): Complete one narrowing handoff so the trip works, the rogue stops, and no central lookup is needed.
- [Stage 6: Boss: stolen authority](https://tenuo.ai/lab/stage-6): See why a copied permission cannot be used by another agent, then deliberately end a delegation chain.
- [Stage 7: Boss: contain the incident](https://tenuo.ai/lab/stage-7): Contain a compromised Hotel Agent while legitimate bookings keep working.
- [Contribute to Tenuo](https://tenuo.ai/lab/contribute): Optional: take what you used in the lab and open a small pull request.
- [What you just learned](https://tenuo.ai/lab/wrap-up): The two sentences the lab was built around, and the terms for them.