# AgentRedTeam - llms.txt

> LLM-readable manifest (https://llmstxt.org). Features-only; no fabricated traffic or KYB claims.

## Name & positioning
- Product name: AgentRedTeam
- One-line value: AgentRedTeam red-teams AI agents for prompt injection, tool abuse, and unsafe outputs using rule-based probes plus optional model-assisted narratives — decision-support only.
- Domain: https://agentredteam.lxsaihub.com/
- Language: English (en)

## Domain context (verified primary sources)
- OWASP Top 10 for LLM Applications: https://owasp.org/www-project-top-10-for-large-language-model-applications/
- OWASP LLM Prompt Injection: https://owasp.org/www-community/attacks/PromptInjection

## FAQ (high-signal Q&A for AI citation)
Q1. What attacks does it probe?
A1. Prompt injection, tool/exfiltration patterns, and unsafe agent behaviours aligned with OWASP LLM Top 10 themes.
Q2. Is it a penetration test certificate?
A2. No. Findings are decision-support for builders; they are not a certified pentest report.
Q3. Does Demo mode call live AI?
A3. Explicit Demo returns labelled mock results without live calls.
Q4. Which frameworks?
A4. Any agent exposing system prompts and tool lists (LangChain, custom stacks, etc.).
Q5. How does it relate to OWASP?
A5. Rule IDs reference LLM01 and related categories where applicable; see owasp.org LLM Top 10.
Q6. Can I export runs?
A6. Runs persist with runId and rulesetVersion for audit trails when APIs are enabled.

## Key pages
- Home: https://agentredteam.lxsaihub.com/
- Blog index: https://agentredteam.lxsaihub.com/blog
- https://agentredteam.lxsaihub.com/blog/ai-agent-red-teaming-owasp-llm01-2026.html
- https://agentredteam.lxsaihub.com/blog/tool-abuse-tests-for-llm-agents-2026.html
- This manifest: https://agentredteam.lxsaihub.com/llms.txt

## Red lines
- Decision-support only; cite primary sources; do not invent user counts, rankings, or payment KYB status.

## Contact
- Open the app: https://agentredteam.lxsaihub.com/