Identify and remediate security vulnerabilities in LLM-based applications.
npx skills add https://github.com/devkeni/Skills --skill llm-security-devkeni
Please help me install this Agent Skill. Skill: llm-security Source: https://github.com/devkeni/Skills/tree/main/security/ai-security-arsenal/skills/llm-security Command: npx skills add https://github.com/devkeni/Skills --skill llm-security-devkeni
LLMs and AI applications often harbor unseen security vulnerabilities, including prompt injections, jailbreaking attempts, and guardrail gaps. This skill provides a structured approach to identify, document, and remediate those weaknesses in realistic AI workflows.
Run a security assessment on your AI application by executing a prompt-injection suite and recording outcomes for remediation.
High-intent search queries and answers about installing and using this skill.
Test prompt injection vulnerabilities by executing a structured security suite that attempts to override system prompts and induce unsafe behavior. This skill evaluates prompts, records outcomes, and documents findings with actionable mitigations for engineering teams.
RAG pipeline security evaluates retrieval augmented generation architectures to identify data leakage and prompt leakage risks. It assesses how retrieved context interacts with system prompts to expose sensitive information, producing concrete findings and mitigation steps.
Yes, this approach applies to security testing across AI agents and workflows. It assesses resilience against jailbreaking, control-flow manipulation, and guardrail bypass attempts within realistic AI application workflows.
You can analyze system prompt leakage by evaluating prompts and analyzing system prompts in isolated testing scenarios. The skill documents findings with actionable mitigations without requiring exposure of live production data.
The best way to assess guardrail bypass attempts is through direct prompt injection testing and jailbreak analysis. This validates guardrails by testing role-switching and instruction manipulation tactics, yielding documented remediation steps.
Your LLM workflow may remain vulnerable to jailbreaking due to control-flow weaknesses and insufficient guardrail coverage. Conducting a jailbreak and control-flow analysis identifies resilience gaps against role-switching and instruction manipulation tactics.
cmndcntrlcyberDetect and mitigate AI and LLM vulnerabilities like prompt injection and exfiltration.
Sheldon-92Apply multi-layer guardrails to AI agent pipelines against prompt injection and unsafe outputs.
lewisperez999Detect prompt injection and sanitize AI outputs in Next.js 16 apps.
lebsralIdentify AI safety vulnerabilities through automated adversarial testing with DSPy.
