Detect prompt injection and agentic tool misuse against OWASP ASI standards.
npx skills add https://github.com/pdparchitect/rook --skill hunt-llm-ai-pdparchitect
Please help me install this Agent Skill. Skill: hunt-llm-ai Source: https://github.com/pdparchitect/rook/tree/main/skills/hunt-llm-ai Command: npx skills add https://github.com/pdparchitect/rook --skill hunt-llm-ai-pdparchitect
This skill addresses the critical security gaps in LLM-backed applications, such as prompt injection, unauthorized data access, and agentic tool misuse, which are often overlooked in standard security audits.
Use the hunt-llm-ai skill to audit the target chatbot endpoint for prompt injection and IDOR vulnerabilities.
High-intent search queries and answers about installing and using this skill.
To audit an LLM chatbot for prompt injection and data exfiltration, you systematically test model inputs and tool-use interfaces. This identifies injection chains and verifies potential impact against OWASP ASI security standards.
Indirect prompt injection in RAG endpoints occurs when malicious instructions are hidden in retrieved documents. It is detected by testing model inputs for document-based injection chains and ASCII smuggling techniques.
To test autonomous AI agents for agentic tool misuse and privilege escalation, evaluate the system against the OWASP ASI 2026 framework. This verifies goal hijacking vulnerabilities and unauthorized tool execution.
Yes, this security audit evaluates AI agents against the OWASP ASI 2026 framework. It systematically tests chatbots, RAG endpoints, and autonomous agents to verify security vulnerabilities and potential impact.
The best way to secure a customer support chatbot with internal database access is auditing its endpoints for prompt injection and IDOR vulnerabilities. This ensures it cannot be manipulated into revealing private user data.
vasilyu1983Audits codebases for undocumented APIs and generates coverage reports.
selfagencyAssess AI agent codebases against OWASP ASI Top 10 security risks.
Zhow01Detect and block unsafe prompts and secret leakage in OpenClaw deployments.
z1439527767Audit AI skills and bundles for security vulnerabilities via static analysis.
