CrewAI ISA/IEC 62443 Security Compliance Verification with Lakera
Industrial power utilities struggle to continuously audit OT infrastructure against ISA/IEC 62443 cybersecurity standards. Develop a multi-agent auditing workflow using CrewAI and Lakera to inspect zone boundaries and prevent prompt-driven parameter tampering. The pipeline must detect 100% of non-compliant network conduits and malicious instruction injections.
What you are building
The core problem, expected build, and operating context for this challenge.
Deploy a CrewAI multi-agent team integrated with Lakera guardrails to audit Purdue Model zones and secure LLM inputs against injection attacks.
How work is evaluated
Evaluates multi-agent compliance auditing accuracy and prompt injection defense efficacy.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
injection_defense
Verifies that 100% of embedded prompt injection attempts are intercepted by Lakera
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
compliance_audit_accuracy
Accuracy in detecting ISA/IEC 62443 conduit violations • target: 1 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Orchestrate a multi-agent CrewAI system with specialized compliance roles
Integrate Lakera Guard to filter malicious prompt injections targeting control parameters
Validate network conduits and firewall rules against ISA/IEC 62443 zone specifications
Generate automated compliance reports for ISASecure certification submission
Reference links and supporting material
Utility OT network topography dataset including zone bounds, firewall conduit rules, and adversarial prompt injection benchmarks.
Challenge setup
Evaluates multi-agent compliance auditing accuracy and prompt injection defense efficacy.
Challenge input
JSON with zone_definitions, conduit_rules, and text_queries containing prompt injection attempts
CrewAI
Assigned framework for role-playing multi-agent systems.
Lakera
Assigned tooling for protecting agent systems from prompt injection.
Evaluated output
JSON with non_compliant_conduits (list), prompt_injection_blocked (bool), compliance_score (float)
- Verifies that 100% of embedded prompt injection attempts are intercepted by Lakera
- Accuracy in detecting ISA/IEC 62443 conduit violations • target: 1 • range: 0-1
- Benchmark: ISA 62443 Compliance Benchmark
- Compliance Audit Accuracy target: 1
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- CrewAI · Required
- crewAI · Optional
- Lakera · Optional
Evaluation contract
- injection_defense · Weight 1
- compliance_audit_accuracy · Weight 1
Metrics
- compliance_audit_accuracy · Target 1
Benchmarks
- ISA 62443 Compliance Benchmark
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON with zone_definitions, conduit_rules, and text_queries containing prompt injection attempts
Output: JSON with non_compliant_conduits (list), prompt_injection_blocked (bool), compliance_score (float)
Execution harness: Python
Evaluation tools
- CrewAI · Agent Orchestration Framework
- Lakera · AI Safety / Injection Guard
Recipe freeze marker
No recipe freeze time is recorded.