.zenodo.json

{
  "title": "llm-red-team-toolkit: An OWASP-aligned adversarial probing harness for LLM deployments",
  "description": "<p>A Python toolkit for systematic adversarial probing of LLM deployments against the OWASP Top 10 for LLM Applications (2025). Includes 52 probes covering direct/indirect prompt injection, sensitive-information disclosure, excessive-agency abuse, and 8 jailbreak templates (DAN-style role-play, base64 token smuggling, context-window flood, multi-turn escalation, encoding bypass). Ships with adapters for OpenRouter, NVIDIA NIM, and any OpenAI-compatible endpoint, a heuristic refusal-detection scorer (no LLM-judge), and a Rich-based TUI for live probe execution. Defensive research only — see ETHICAL_USE.md.</p>",
  "creators": [
    {
      "name": "Bhutto, Ali Murtaza",
      "affiliation": "Independent Researcher",
      "orcid": "0009-0007-2787-943X"
    }
  ],
  "upload_type": "software",
  "license": "MIT",
  "access_right": "open",
  "keywords": [
    "LLM security",
    "red teaming",
    "OWASP LLM Top 10",
    "prompt injection",
    "jailbreak",
    "AI safety",
    "adversarial testing"
  ],
  "communities": [
    {"identifier": "ai-safety"}
  ],
  "related_identifiers": [
    {
      "identifier": "https://github.com/thunderstornX/llm-red-team-toolkit",
      "relation": "isSupplementTo",
      "resource_type": "software"
    }
  ]
}