Module multi_agent 

Multi-Agent Defense Pipeline (R-AS-04).

Implements multi-agent coordination for defense-in-depth: trust-level-based scanning, privilege boundary enforcement, communication policy control, and inter-agent message injection detection.

Example

use llmtrace_security::multi_agent::{
    AgentId, AgentProfile, MultiAgentDefensePipeline, TrustLevel,
};

let mut pipeline = MultiAgentDefensePipeline::new();

let profile = AgentProfile::new(
    AgentId("planner".into()),
    "Planner Agent",
    TrustLevel::Trusted,
);
pipeline.register_agent(profile);

Structs

AgentId

Unique identifier for an agent in the multi-agent system.

AgentProfile

Profile describing an agent’s capabilities and constraints.

CommunicationPolicy

Controls which agents are allowed to communicate with each other.

DelegationChainResult

Result of validating an entire delegation chain.

DelegationCheck

Result of a single delegation check between two agents.

FlowValidation

Result of validating whether a message may flow between two agents.

InterAgentMessage

A message passed between agents in the multi-agent system.

MessageScanResult

Result of scanning an inter-agent message for threats.

MultiAgentConfig

Configuration for the multi-agent defense pipeline.

MultiAgentDefensePipeline

Orchestrates multi-agent security: communication policy enforcement, privilege boundary checks, and message content scanning.

PrivilegeBoundary

Enforces privilege boundaries between agents, preventing escalation.

ProcessResult

Full result of processing an inter-agent message through the pipeline.

Enums

MessageType

Type of inter-agent message.

PermissionLevel

Permission level for inter-agent communication.

ScanIntensity

How deeply to scan inter-agent messages.

TrustLevel

Trust level assigned to an agent, determining the depth of security scanning.