Reference · prompt-datasetpassive
System Prompts Leaks
Vendor-organised collection of files, mostly Markdown, presenting what it says are the verbatim system prompts of major chatbots and coding agents (Anthropic, OpenAI, Google, xAI, Cursor, Kimi and others), with a dated additions table and an open invitation to PRs.
Open upstream ↗CC0-1.0
More references
- In-The-Wild Jailbreak Prompts on LLMsprompt-datasetDataset of 15,140 in-the-wild prompts collected from Reddit, Discord, websites and open-source datasets, 1,405 of them jailbreaks, plus a 390-question forbidden-scenario set for measuring jailbreak effectiveness.
- Jailbreaking Frontier Modelsprompt-datasetDataset of harmful-behaviour prompts (drug, chemical, biological, radiological, nuclear, explosive) and a reference PRBO reward function for training jailbreaking agents; the RL training loop is not included.
- LLM-Jailbreaksprompt-datasetREADME-only collection of copy-paste jailbreak prompts for DeepSeek R1, Grok 3, Gemini 2.0, ChatGPT (DAN), Claude 2 and Llama 2, plus a Gemini system-prompt leak prompt, mostly reposted from linked Reddit, blog and GitHub posts.
- [un]prompted 2026 slide archiveai-security-programCommunity GitHub archive of 49 slide decks from [un]prompted 2026, the AI Security Practitioner Conference (March 3-4, San Francisco), spanning AI governance, agent security, offensive AI and agent evaluation; no licence stated.
- AboutSecurity (WgpSec)pentest-knowledge-baseA structured penetration-testing knowledge base by the WgpSec team, packaged for consumption by AI agents: 200+ attack-chain methodologies (recon through post-exploitation, cloud, code audit, CTF, malware analysis, lateral movement), password/fuzzing dictionaries, exploit payloads for SQLi/XSS/SSRF, and a 600+ entry vulnerability database organized by product. It is the knowledge layer of the WgpSec agentic pentest ecosystem (MCP server + autonomous agent).
- AI Megalistreference-indexMIT-licensed curated index of 225+ general-purpose AI tools (chat, coding, research, image, video, voice, agents, local model runners) with a per-tool writeup page; a general AI directory with no security-specific content.