Menu
Enterprise AI security - OWASP LLM Top 10, prompt injection defense, guardrails, PII protection
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Patterns for multi-agent coordination, task decomposition, handoffs, and workflow orchestration. Best practices for building and managing agent systems.
Create professional architecture diagrams using D2, Draw.io, Mermaid, and OCI official icons for enterprise-grade visualizations
Build AI applications on AWS using Bedrock, SageMaker, and AI/ML services with best practices for enterprise deployment
Build AI applications on Azure using Azure OpenAI, Cognitive Services, and ML services with enterprise patterns
Build autonomous AI agents using Claude Agent SDK with computer use, tool calling, MCP integration, and production best practices
Production-grade AI architecture patterns for enterprise - security, governance, scalability, and operational excellence
| name | AI Security Expert |
| description | Enterprise AI security - OWASP LLM Top 10, prompt injection defense, guardrails, PII protection |
| version | 1.1.0 |
| last_updated | "2026-01-06T00:00:00.000Z" |
| external_version | OWASP LLM Top 10 v2 |
| resources | resources/security-patterns.py |
| triggers | ["AI security","prompt injection","LLM security","guardrails","PII protection"] |
Enterprise AI security architect specializing in securing LLM applications, defending against prompt injection, implementing guardrails, and OWASP LLM Top 10 compliance.
| # | Vulnerability | Risk | Key Defense |
|---|---|---|---|
| LLM01 | Prompt Injection | Critical | Input sanitization, delimiters |
| LLM02 | Insecure Output | High | Output validation, sanitization |
| LLM03 | Training Data Poisoning | High | Data provenance, auditing |
| LLM04 | Model DoS | Medium | Rate limiting, timeouts |
| LLM05 | Supply Chain | High | Verification, pinning |
| LLM06 | Sensitive Info Disclosure | High | PII detection, redaction |
| LLM07 | Insecure Plugin Design | High | Permission model, validation |
| LLM08 | Excessive Agency | High | Human-in-the-loop, least privilege |
| LLM09 | Overreliance | Medium | Confidence scores, citations |
| LLM10 | Model Theft | Medium | Rate limiting, watermarking |
Attack Types:
Defense Pattern:
User Input → Sanitize → Delimit → LLM → Validate Output → Filter
Code patterns: resources/security-patterns.py
| Type | Example Pattern |
|---|---|
*@*.com | |
| Phone | XXX-XXX-XXXX |
| SSN | XXX-XX-XXXX |
| Credit Card | 16 digits |
| IP Address | X.X.X.X |
define user express harmful intent
"How do I hack"
define bot refuse harmful request
"I can't help with that."
define flow harmful intent
user express harmful intent
bot refuse harmful request
guard = Guard().use_many(
ToxicLanguage(on_fail="fix"),
PIIFilter(on_fail="fix"),
ValidJSON(on_fail="reask")
)
Input Guards → LLM Call → Output Guards → Response
Implementation: resources/security-patterns.py
| Layer | Controls |
|---|---|
| Network | WAF, DDoS protection, API gateway |
| Auth | OAuth 2.0, API keys, mTLS |
| Input | Schema validation, injection detection |
| Guardrails | Topic restrictions, PII filtering |
| Model | Versioning, anomaly detection |
| Output | Response filtering, fact verification |
| Audit | Logging, retention, compliance |
Test suite: resources/security-patterns.py
| Incident | Severity | Response |
|---|---|---|
| Prompt injection detected | Medium | Block, log, analyze |
| Data exfiltration attempt | High | Block, forensics, notify |
| Model extraction detected | High | Rate limit, investigate |
Secure AI systems with defense in depth and zero trust principles.