skills.rest
jpoindexter
jpoindexterCOMMUNITY
Explore all skills by this author →
Website𝕏@microninkGitHub

llm-safety-and-multitenancy

Guide secure LLM engineering with cross-tenant isolation strategies.

1|Updated Jun 9, 2026
Search Tags:#security engineering#llm safety#data leak prevention#multi-tenant isolation#untrusted input
One-click install
npx skills add https://github.com/jpoindexter/design-and-ai-skills --skill llm-safety-and-multitenancy
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: llm-safety-and-multitenancy
Source: https://github.com/jpoindexter/design-and-ai-skills/tree/main/ai-engineering-skills/llm-safety-and-multitenancy
Command: npx skills add https://github.com/jpoindexter/design-and-ai-skills --skill llm-safety-and-multitenancy

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pypdf2, beautifulsoup4, lxml, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill Unit offers a robust guide for engineers and security teams on LLM safety and multi-tenant isolation, addressing the threats of untrusted-input attacks and preventing data leaks.

Core Features & Use Cases

  • LLM Safety: Understand and mitigate untrusted-input threats following the OWASP LLM Top 10.
  • Multi-Tenant Isolation: Implement controls for data-leakage prevention, permission boundaries, and cross-tenant isolation.
  • Use Case: For a multi-tenant chatbot service, ensure secure handling of user sessions and prevent data exposure across different customer contexts.

Quick Start

Run the 'build-skills-index.py' script located in the 'scripts/' directory to compile the LLM Safety & Multi-Tenant Isolation skill index into the 'ai-engineering-skills/index.html' file.

Frequently Asked Questions about llm-safety-and-multitenancy

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I prevent data leakage in multi-tenant LLM systems?▼

Prevent data leakage in multi-tenant LLM systems by implementing cross-tenant isolation controls and strict permission boundaries. This Skill provides comprehensive engineering guidance to secure user sessions and prevent data exposure across different customer contexts.

What is the OWASP LLM Top 10 untrusted-input threat model?▼

The OWASP LLM Top 10 untrusted-input threat model defines security vulnerabilities where malicious user prompts manipulate model behavior. This Skill explains these LLM safety threats and provides mitigation strategies to protect your multi-tenant chatbot service.

How to build cross-tenant isolation for a multi-tenant chatbot service?▼

Build cross-tenant isolation for a multi-tenant chatbot service by enforcing permission boundaries and secure session handling. This Skill outlines specific security engineering strategies to prevent unauthorized data exposure across different customer contexts.

Can I use Python and BeautifulSoup to compile LLM security guidelines?▼

Yes, you can use Python and BeautifulSoup to compile LLM security guidelines by running the provided build-skills-index.py script. This script requires pypdf2 and lxml dependencies to generate an HTML skill index for your LLM safety reference.

What are the limitations of multi-tenant isolation for untrusted input?▼

Limitations of multi-tenant isolation for untrusted input include the persistent risk of cross-tenant data leakage if permission boundaries are misconfigured. This Skill addresses these constraints by mapping comprehensive security controls against the untrusted-input threat model.

Related Skills

Gimpy666Gimpy666
1

scraped_docs

Convert Tribot JavaDoc HTML pages into per-class Markdown files and a consolidated reference.

Community
Intermediate
Camilo8902Camilo8902

multi-tenant-safety-checker

Enforce tenant isolation in PostgreSQL with Row Level Security policies.

Community
Advanced
hiroto0701hiroto0701

security-patterns

Implement JWT authentication and security patterns for hardened applications.

Community
Advanced
maruakshaymaruakshay
4

rag-security

Audit RAG systems for ingestion poisoning, retrieval boundary failures, and metadata weaknesses.

Community
Advanced