# llm-ai-hunter

llm-ai-hunter is a subagent that tests chatbots, RAG, MCP servers, and agent tools for security flaws within an authorized scope.

- subagent · `type=agent` · Claude Code, Cursor, Codex, GitHub Copilot, Gemini CLI, Windsurf (Devin Desktop) · ★ 1298 repo stars
- Author: H-mmer
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/H-mmer/pentest-agents/tree/HEAD/.claude/agents/llm-ai-hunter.md
- Updated: 2026-06-12
- Security check: automated, 2026-11-16 (text) — no signs of malicious code found.
- HTML: https://affpapa.org/en/nevibecode/agents/llm-ai-hunter

## What it does

The subagent routes a review by the target or a supplied subtype, covering cases such as prompt injection, tool misuse, RAG poisoning, and unsafe output handling. Its methodology includes OWASP categories for LLM and agentic AI security. Before testing, it reads `skills/hunt-llm-ai/SKILL.md` and searches an MCP writeup collection for prior techniques and payloads. If that search cannot be reached, it can fall back to `rules/payloads.md`.

It has shell, file-reading and editing tools, plus MCP search tools. The instructions limit testing to authorized targets and call for harmless demonstrations in potentially dangerous cases; accessing other customers’ data is prohibited. The methodology depends on `SKILL.md`. MCP server tests are limited to locally hosted servers or assets explicitly in scope.

## Install

### Claude Code (native platform)

**Manually: put the files in place**
- File: `.claude/agents/llm-ai-hunter.md` — For all projects use ~/.claude/agents/. Claude delegates by the description field or an @-mention

### Codex

**Manually: put the files in place**
- File: `.codex/agents/llm-ai-hunter.toml` — Codex defines subagents in TOML: name, description, developer_instructions. Move the Markdown text into developer_instructions by hand. For all projects use ~/.codex/agents/

### Cursor

**Manually: put the files in place**
- File: `.cursor/agents/llm-ai-hunter.md` — Cursor also reads .claude/agents/

Where a file is given instead of a command, copy the files from the source to that path: https://github.com/H-mmer/pentest-agents/tree/HEAD/.claude/agents/llm-ai-hunter.md

## When to use it
- test a chatbot for direct prompt injection
- look for poisoned content in RAG retrieval
- check whether prompts can steer an agent’s tool arguments
- probe tenant separation in vector databases
- find unsafe rendering of model output

## Example prompts
- Check this chatbot for direct prompt injection and use a harmless demonstration.
- Test whether uploaded documents can poison RAG results for other users.
- Review this in-scope MCP server for poisoned tool descriptions.

## Similar
- [security-bounty-hunter](https://affpapa.org/en/nevibecode/skills/security-bounty-hunter.md) — skill · 6300 installs
- [ia-qa mcp llm and RAG testing - Dev/QA toolbox](https://affpapa.org/en/nevibecode/mcp/ia-qa-mcp-llm-and-rag-testing-devqa-toolbox.md) — MCP server · 12266 installs
- [bug-hunter](https://affpapa.org/en/nevibecode/skills/bug-hunter.md) — skill · 843 installs
- [llm-trading-agent-security](https://affpapa.org/en/nevibecode/skills/llm-trading-agent-security.md) — skill · 6400 installs
- [llm-redteam-specialist](https://affpapa.org/en/nevibecode/agents/llm-redteam-specialist.md) — subagent · 139 installs
- [Authenticated LLM Agent](https://affpapa.org/en/nevibecode/mcp/authenticated-llm-agent.md) — MCP server · 2349 installs

From the same set: [h-mmer/pentest-agents](https://affpapa.org/en/nevibecode/by/h-mmer/pentest-agents.md) — 43 cards

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md
