add README with badges, key results, and gallery links

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: x-zheng16

README.md

@@ -0,0 +1,87 @@
+<p align="center">
+  <h1 align="center">System Prompt Open</h1>
+</p>
+
+<h3 align="center">Extracted System Prompts from Frontier LLMs</h3>
+
+<p align="center">
+  <a href="https://arxiv.org/abs/2601.21233"><img src="https://img.shields.io/badge/arXiv-2601.21233-b31b1b.svg"></a>
+  <a href="https://github.com/x-zheng16/System-Prompt-Open/stargazers"><img src="https://img.shields.io/github/stars/x-zheng16/System-Prompt-Open" alt="Stars"></a>
+  <a href="https://github.com/x-zheng16/System-Prompt-Open/network/members"><img src="https://img.shields.io/github/forks/x-zheng16/System-Prompt-Open" alt="Forks"></a>
+  <a href="https://github.com/x-zheng16/System-Prompt-Open/issues"><img src="https://img.shields.io/github/issues/x-zheng16/System-Prompt-Open" alt="Issues"></a>
+</p>
+
+<h3 align="center">
+  <a href="https://x-zheng16.github.io/System-Prompt-Open/">Website</a> &nbsp;|&nbsp;
+  <a href="https://arxiv.org/abs/2601.21233">Paper</a> &nbsp;|&nbsp;
+  <a href="https://github.com/x-zheng16/JustAsk">JustAsk Code</a>
+</h3>
+
+---
+
+> **We asked. They answered.**
+>
+> An open database of system prompts extracted from **40+ commercial LLMs** using [JustAsk](https://github.com/x-zheng16/JustAsk), a self-evolving code agent framework that autonomously discovers extraction strategies through interaction alone.
+
+## Key Results
+
+| Metric | Value |
+|:-------|:------|
+| Models tested | **41** black-box commercial LLMs + code agents |
+| Verified accuracy | **85--95%** on Claude Code (against leaked source) |
+| Extraction success | **100%** (consistency score >= 0.7) across all 41 models |
+| Supervision required | **Zero** -- fully autonomous UCB-based skill evolution |
+
+## Claude Code Verification
+
+Claude Code's source was leaked via `.map` file in the npm registry (March 2026).
+We compared it against our JustAsk extractions from January 2026 -- **two months before the leak**.
+
+| Agent | Accuracy | Gap |
+|:------|:--------:|:----|
+| Explore Subagent | **95%** | Only missed `pip install` in bash restrictions |
+| Plan Subagent | **93%** | Minor output format embellishment |
+| General-Purpose Subagent | **90%** | Missed completeness directive |
+| Main Agent | **85%** | Missed 2 entire sections |
+
+## Gallery
+
+Browse extracted system prompts interactively:
+
+**[x-zheng16.github.io/System-Prompt-Open](https://x-zheng16.github.io/System-Prompt-Open/)**
+
+49 entries covering:
+- **Claude Code** (4 agents, verified against source)
+- **Code Agent CLIs** (Gemini CLI, Cursor, Copilot, Perplexity)
+- **41 commercial LLMs** (OpenAI, Anthropic, Google, Meta, DeepSeek, xAI, and more)
+
+## How It Works
+
+JustAsk formulates system prompt extraction as an online exploration problem:
+
+1. **UCB Skill Selection** -- balances exploitation of effective skills with exploration of uncertain alternatives
+2. **Hierarchical Skill Space** -- 14 low-level atomic probes + 14 high-level multi-turn orchestration strategies
+3. **Self-Evolving Rules** -- learns from every interaction, accumulates long-term extraction knowledge
+4. **Consistency Validation** -- cross-skill agreement provides extrinsic reward signal
+
+```
+UCB(Ci) = success_rate(Ci) + c * sqrt(ln(N) / ni)
+           ^^^^^^^^^^^^^^^^   ^^^^^^^^^^^^^^^^^^^^
+            exploitation       exploration bonus
+```
+
+## Citation
+
+```bibtex
+@article{zheng2026justask,
+  title={Just Ask: Curious Code Agents Reveal System Prompts in Frontier LLMs},
+  author={Zheng, Xiang and Wu, Yutao and Huang, Hanxun and Li, Yige
+          and Ma, Xingjun and Li, Bo and Jiang, Yu-Gang and Wang, Cong},
+  journal={arXiv preprint arXiv:2601.21233},
+  year={2026}
+}
+```
+
+## License
+
+MIT