Automate Data Center Security Audit Workflows with OpenAI Agents SDK and Sourcegraph Amp
India's rapid expansion of hyperscale data centers introduces physical and cybersecurity monitoring blind spots. In this challenge, you will implement an autonomous compliance and threat audit agent using the OpenAI Agents SDK and Sourcegraph Amp. The agent will automatically scan system logs, network telemetry, and physical access logs across distributed data center facilities, detect security posture anomalies, and trigger automated remediation workflows while refactoring infrastructure codebase policies.
What you are building
The core problem, expected build, and operating context for this challenge.
Build multi-turn security auditing agents using OpenAI Agents SDK alongside Sourcegraph Amp code refactoring integration.
How work is evaluated
Evaluates the detection of unauthorized access events and code vulnerability patches.
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.
threat_detection_test
Ensures physical override anomaly is flagged as critical.
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
recall
Percentage of security threats correctly identified. • 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
Implement OpenAI Agents SDK runner loops with structured tool function calling.
Leverage Sourcegraph Amp for code-level security patch generation.
Analyze data center network access telemetry for zero-day unauthorized entry vectors.
Generate automated compliance reports mapped to ISO 27001 and CERT-In advisories.
Reference links and supporting material
Data center physical security badge logs merged with network edge firewall event records.
Challenge setup
Evaluates the detection of unauthorized access events and code vulnerability patches.
Challenge input
JSON array of security logs containing timestamps, IP addresses, facility zones, and authorization status.
OpenAI Agents SDK
Required SDK for autonomous function-calling loops.
Sourcegraph Amp
Required tool for codebase search and security patch recommendation.
Evaluated output
JSON object with detected_threats, vulnerability_severity, and patch_recommendation.
- Ensures physical override anomaly is flagged as critical.
- Percentage of security threats correctly identified. • target: 1 • range: 0-1
- Benchmark: Cyber-Physical Security Benchmark
- Recall 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
- Sourcegraph Amp · Required
- OpenAI · Optional
- Cartesia · Optional
- Sourcegraph Amp · Required
Evaluation contract
- threat_detection_test · Weight 1
- recall · Weight 1
Metrics
- recall · Target 1
Benchmarks
- Cyber-Physical Security Benchmark
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: JSON array of security logs containing timestamps, IP addresses, facility zones, and authorization status.
Output: JSON object with detected_threats, vulnerability_severity, and patch_recommendation.
Execution harness: Python
Evaluation tools
- OpenAI Agents SDK · Agent Execution Framework
- Sourcegraph Amp · Code Audit Tool
Recipe freeze marker
No recipe freeze time is recorded.