Back to Ecosystem Pulse
EvaluationProduction Ready
T3MP3ST
by elder-plinius
Autonomous multi-agent red-teaming harness for offensive security testing
TypeScript
Updated Aug 24, 2026
Share:
What It Does
Runs autonomous red-team campaigns using configurable multi-agent attackers to probe system defenses. Event-Driven Agent Pattern Orchestrates attacker agents, blue-team defenders, and mission logic to simulate realistic adversarial workflows and find gaps. Orchestrator-Worker Pattern Provides harness features for repeatable scenarios, logging, and automated vulnerability discovery at scale.
Why It Matters
As agent-driven systems proliferate, you need ways to surface failure modes under adversarial conditions; simple unit tests won't reveal attacker strategies. T3MP3ST lets teams stress-test agents and orchestration by simulating coordinated offensive campaigns, producing reproducible signals about agent reliability and weak points. That adversarial evaluation is essential to build agent track records and trust before deployment. See how Supply Chain Compromise can impact integrity in multi-agent deployments.
Ideal For
Security teams and platform engineers who need repeatable, adversarial evaluation of multi-agent systems and agent workflows. This approach aligns with the Agent Service Mesh Pattern to manage complex agent interactions at scale.
Use Cases
- Stress-testing agent orchestration with coordinated attacker scenarios
- Finding emergent agent failure modes through adversarial campaigns
- Generating reproducible logs and traces for agent reliability assessments
- Running regression-style red-team tests as part of pre-production checks
Works With
openai
Topics
agentsaimulti-agentoffensive-securityredteam
Similar Tools
agent-arenaagent-playground
Keywords
multi-agent orchestrationcontinuous agent evaluationadversarial testingagent failure modes