Dayward AIWeChat Mini Program
All courses

5 days · about 23 hours

Agent Security in 5 Days

Take one internal assistant that happens to hold all three parts of the lethal trifecta and turn it into a target: five days from threat modeling to architectural defense, runtime controls and a red-team loop that drives attack success to zero while keeping the agent useful.

Sign in and link WeChat to share one free article per day across both platforms. Unlock more with rewarded videos in the mini program.

Start day one

By the end

A security report carrying two numbers: attack success rate driven from high to zero, task completion barely moved, and every defense able to say which class of attack it actually stops

One fictional internal knowledge assistant runs through all five days: first map its trust boundaries, then work the input side, the architecture and the runtime in turn, and finally fold everything into a red-team loop you can keep running

View account and linked status →5 days · about 23 hours

Week 1

From seeing the threat to stopping it

01
D1🔒 Locked

Threat modeling: the lethal trifecta, trust boundaries, and how to use two OWASP lists as checklists

Start with the lethal trifecta test to see where an agent is actually dangerous: once private data, untrusted content and outbound communication coexist, one successful prompt injection becomes one data exfiltration. Then draw it as a trust boundary diagram and walk both OWASP lists against it.

02
D2🔒 Locked

An injection range and the input side: what four tiers of defense stop, and where each one fails

Build a reproducible injection range, bolt on delimiters, provenance markers, spotlighting and a detector one tier at a time, and use attack success rate together with task completion to see exactly how far each tier gets — and why none of them is a boundary.

03
D3🔒 Locked

Architectural defense: six design patterns, and how plan-then-execute drives attack success to zero

Move the defense out of the prompt and into the structure: what action-selector, plan-then-execute, map-reduce, dual LLM, code-then-execute and context-minimization each constrain, then land two of them on the range so attack success hits zero while the agent keeps working.

04
D4🔒 Locked

Runtime controls: capabilities, approval gates, egress allowlists, sandbox layers, secrets and tenancy

Assume the injection already happened and use the runtime to contain the damage: declare each tool's capability and risk tier, gate the dangerous ones on human approval, allowlist filesystem paths and network egress, separate the policy layer from the isolation layer, and handle secrets, per-user credentials and tenant isolation.

05
D5🔒 Locked

Red teaming: keeping an attack suite alive, wiring it into CI, and what to do after an incident

Turn four days of defenses into a loop you can keep running: maintain an attack suite that grows, write a red-team script that produces a report, gate CI on both numbers, and add rogue-behavior detection, a kill switch and a minimal incident response plan.