Lobster Trap Integration

Veea

Export AgentShield rules as Lobster Trap YAML policies for runtime enforcement

How It Works
AgentShield
Install-time scanning
Config auditing
Output scanning
+
Lobster Trap
Runtime DPI proxy
Prompt inspection
Policy enforcement
=
Full Lifecycle
Scan before install
Enforce at runtime
Audit everything

GATE scans skills before they activate → generates policy rules

Lobster Trap enforces those rules at runtime on every prompt/response

Together — no malicious skill installs AND no runtime prompt injection

Quick Setup
1
Generate policy above
AgentShield exports its 33+ threat rules as Lobster Trap conditions
2
Save as policy.yaml
Copy the YAML and save it alongside the Lobster Trap binary
3
Run Lobster Trap
./lobstertrap serve --backend http://localhost:11434 --policy policy.yaml
4
Point agents at the proxy
Agents connect to localhost:8080instead of the LLM directly. Lobster Trap inspects every prompt and response using AgentShield's rules.