Skip to main content
Loading navigation
Back to challenges

Supply Chain Resilience Swarm Agents

Recent supply chain attacks on tools like LiteLLM highlight the need for automated incident response. In this challenge, you will use CrewAI to orchestrate a team of agents that monitor data pipelines and react to security breaches. You will utilize Protecto for data privacy masking within agent communications and Agency for real-time observability of the agent swarm. The system will leverage the GPT-5.3-Codex model for its core reasoning engine, ensuring that incidents are handled with high precision and logical consistency.

Workflow AutomationHosted by Vera
Sign in free to start

What you are building

Build a multi-agent swarm with CrewAI and GPT-5.3-Codex to monitor and respond to supply chain security incidents.

How work is evaluated

Evaluation

Evaluation centers on the swarm's ability to triage a mock supply chain incident and correctly delegate tasks.

Datasets

Public datasets, plus the private datasets you have access to.

How submissions are scored

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

PII Masking Check

Agent communications must not contain unmasked sensitive data.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

Delegation Efficiency

Percentage of tasks correctly routed to the specialized agent on the first attempt • target: 90 • range: 0-100

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

  • Master CrewAI role-playing agent teams with specialized agents for incident triage, forensic analysis, and containment

  • Integrate Protecto to ensure that sensitive company data is masked before being processed by the GPT-5.3-Codex reasoning engine

  • Deploy Agentic Dev frameworks to accelerate the creation of custom agent tools for pipeline management

  • Implement Agency (Observability) to track agent thoughts and tool executions in a production-like environment

  • Leverage Ellipsis to provide an interface for human operators to approve agent-suggested containment actions

  • Design complex task dependencies in CrewAI that ensure forensic evidence is gathered before remediation begins

Resources

Dataset notes

Sample data for 1 tasks

Challenge setup

Evaluation centers on the swarm's ability to triage a mock supply chain incident and correctly delegate tasks.

Challenge input

Log stream containing a mix of normal and malicious activity

CrewAI

Framework for orchestrating

Protecto

AI data privacy

Evaluated output

Categorized incident report with severity score

Evaluation checks
  • Agent communications must not contain unmasked sensitive data.
  • Percentage of tasks correctly routed to the specialized agent on the first attempt • target: 90 • range: 0-100
Configured evaluation evidence
  • Delegation Efficiency target: 90
  • 1 public reference case
Configured runtime
  • Docker execution harness

Shows the first task, up to three tools, and up to three evidence items.

View technical recipe

Configured tools

Action Space
  • CrewAI · Required
  • crewAI · Optional
  • Protecto · Optional

Evaluation contract

  • PII Masking Check · Weight 1
  • Delegation Efficiency · Weight 1

Metrics

  • Delegation Efficiency · Target 90

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: Log stream containing a mix of normal and malicious activity

    Output: Categorized incident report with severity score

    Execution harness: Docker

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Supply Chain Resilience Swarm Agents