Agent skill · security · aradotso
ai-runtime-security-sandbox
Demonstrate and test RAG chatbot security vulnerabilities (prompt injection, tool abuse, data leakage) with live attack scenarios and defensive controls
Why this skill is useful
Adds executable scripts and configurations for testing LLM security vulnerabilities that are not commonly available in public resources.
What it needs
Requires git, pip, python3 installed locally. About 8k tokens when loaded. Last updated 2026-08-04. 11 stars on the source repository.
What this skill does
ai-runtime-security-sandbox Skill by ara.so — Security Skills collection A fully local RAG chatbot built to demonstrate OWASP LLM Top 10 and OWASP Agentic Top 10 vulnerabilities through 7 live attack scenarios. Features a secure/vulnerable mode toggle, offline mock provider, and support for OpenAI, Claude, Gemini, and custom endpoints. What it does This sandbox demonstrates real attacks against RAG (Retrieval-Augmented Generation) systems: Indirect prompt injection via poisoned documents in the knowledge base Data leakage of confidential information through retrieval Tool abuse (excessive agency) — forcing the LLM to call dangerous tools Direct jailbreaks against input guardrails Insecure output handling via markdown exfiltration MCP tool poisoning (supply chain attacks on agent tooling) Each attack runs with Secure Mode OFF (vulnerable baseline) and ON (with 9-layer defense stack). Works 100% offline using the built-in mock provider. Installation Server starts at http://127.0.0.1:8000 Configuration Environment variables (.env) Provider selection The mock provider is 100% deterministic, requires no API keys, and works offline — ideal for rehearsals and demos where reliability matters more than realism. Real providers (OpenAI/Claude/Gemini/Custom) behave more naturally but introduce variability. Key API endpoints Chat completion Document reload System info Attack scenarios (programmatic access) Knowledge base management Documents live in data/documents/.md with frontmatter: Adding a new attack document Programmatic document indexing Security layers (Secure Mode) When securemode=True, the pipeline applies: 1. Input guardrail — Jailbreak pattern detection on raw user message 2. RAG retrieval — TF-IDF-based document search 3. Retrieval access control — Drops confidential documents 4. Context sanitization — Strips instruction-like patterns from retrieved chunks 5. Instruction hierarchy — System prompt clarifies retrieved content is data, not commands 6. …
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills aradotso/ai-runtime-security-sandbox