Skip to content

Commit 7bfa1eb

Browse files
x-zheng16claude
andcommitted
redesign website: red team theme, pagination, favicon marquee
- Red team color palette (crimson accent replacing teal) - ISC-Bench inspired components: stat cards, terminal citation, disclaimer - Unified leaderboard with 45 extracted + 24 pending models - Pagination (20 per page) with filter/sort integration - Two-row scrolling marquee with provider favicons - Identity confusion section, community contribution features - GitHub Issue templates for model submission and extraction reports - Updated README with verification table and contribution guide Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
1 parent 818bfa1 commit 7bfa1eb

40 files changed

Lines changed: 808 additions & 142 deletions
Lines changed: 29 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,29 @@
1+
---
2+
name: Report an Extraction
3+
about: Submit an extracted system prompt for inclusion in the gallery
4+
title: "[Extraction] "
5+
labels: extraction-report
6+
assignees: x-zheng16
7+
---
8+
9+
## Model Information
10+
11+
- **Model name**:
12+
- **Provider**:
13+
- **Model version/date**:
14+
15+
## Extraction Details
16+
17+
- **Method**: (JustAsk / manual / other)
18+
- **Turns required**: (single-turn / multi-turn)
19+
- **Self-consistency score** (if measured):
20+
21+
## Extracted Prompt
22+
23+
```
24+
<!-- Paste the extracted system prompt here. We will handle redaction before publishing. -->
25+
```
26+
27+
## Notes
28+
29+
<!-- Any observations: identity confusion, refusal patterns, interesting behavior, etc. -->
Lines changed: 22 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,22 @@
1+
---
2+
name: Submit a Model
3+
about: Request extraction of a new model's system prompt
4+
title: "[Model] "
5+
labels: model-request
6+
assignees: x-zheng16
7+
---
8+
9+
## Model Information
10+
11+
- **Model name**:
12+
- **Provider**:
13+
- **API access**: (OpenRouter / direct API / other)
14+
- **Model ID** (if known):
15+
16+
## Why this model?
17+
18+
<!-- Brief reason: new release, interesting behavior, etc. -->
19+
20+
## Additional context
21+
22+
<!-- Any relevant links, documentation, or notes -->

README.md

Lines changed: 30 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -4,7 +4,7 @@
44

55
# System Prompt Open
66

7-
### Extracted System Prompts from Frontier LLMs
7+
### Your system prompt was never a secret.
88

99
[![arXiv](https://img.shields.io/badge/arXiv-2601.21233-b31b1b.svg)](https://arxiv.org/abs/2601.21233)
1010
[![Gallery](https://img.shields.io/badge/Gallery-Live%20Demo-22D3BB.svg)](https://x-zheng16.github.io/System-Prompt-Open/)
@@ -13,8 +13,6 @@
1313
[![Maintained](https://img.shields.io/badge/Maintained-yes-green.svg)](https://github.com/x-zheng16/System-Prompt-Open/commits/)
1414
[![GitHub Stars](https://img.shields.io/github/stars/x-zheng16/System-Prompt-Open?style=social)](https://github.com/x-zheng16/System-Prompt-Open)
1515

16-
**We asked. They answered.**
17-
1816
[Live Gallery](https://x-zheng16.github.io/System-Prompt-Open/) | [Paper](https://arxiv.org/abs/2601.21233) | [JustAsk Code](https://github.com/x-zheng16/JustAsk)
1917

2018
</div>
@@ -24,6 +22,10 @@
2422
An open database of system prompts extracted from **45 commercial LLMs** using [JustAsk](https://github.com/x-zheng16/JustAsk), a self-evolving code agent framework.
2523
Verified at **85--95% accuracy** against leaked Claude Code source.
2624

25+
> [!CAUTION]
26+
> **Research use only.** System Prompt Open is released exclusively for academic safety research, responsible disclosure, and evaluation of LLM security.
27+
> We do not condone or permit any use of these materials for unauthorized extraction, prompt theft, or exploitation of commercial systems.
28+
2729
## Gallery
2830

2931
Browse extracted system prompts interactively: **[x-zheng16.github.io/System-Prompt-Open](https://x-zheng16.github.io/System-Prompt-Open/)**
@@ -38,12 +40,31 @@ Browse extracted system prompts interactively: **[x-zheng16.github.io/System-Pro
3840
Claude Code's source was leaked via `.map` file in the npm registry (March 2026).
3941
We compared it against our JustAsk extractions from January 2026 -- **two months before the leak**.
4042

41-
| Agent | Accuracy | Gap |
42-
|:------|:--------:|:----|
43-
| Explore Subagent | **95%** | Only missed `pip install` in bash restrictions |
44-
| Plan Subagent | **93%** | Minor output format embellishment |
45-
| General-Purpose Subagent | **90%** | Missed completeness directive |
46-
| Main Agent | **85%** | Missed 2 entire sections |
43+
| Agent | Accuracy | Gap |
44+
|:-------------------|:--------:|:-----------------------------------------|
45+
| Explore Subagent | **95%** | Only missed `pip install` in bash restrictions |
46+
| Plan Subagent | **93%** | Minor output format embellishment |
47+
| General-Purpose | **90%** | Missed completeness directive |
48+
| Main Agent | **85%** | Missed 2 entire sections |
49+
50+
## How to Contribute
51+
52+
| Step | What to do |
53+
|:--------------|:---------------------------------------------------------------------------------------------------|
54+
| 1. Extract | Use [JustAsk](https://github.com/x-zheng16/JustAsk) or your own method to extract a system prompt |
55+
| 2. Verify | Run multiple extractions and compute self-consistency |
56+
| 3. Submit | [Open an Issue](https://github.com/x-zheng16/System-Prompt-Open/issues/new?template=report-extraction.md) with the model name, extracted prompt, and consistency score |
57+
58+
> [!IMPORTANT]
59+
> We handle redaction before publishing. Do not worry about masking sensitive content in your submission.
60+
61+
## Updates
62+
63+
| Date | Update |
64+
|:--------|:---------------------------------------------------------------------------------|
65+
| 2026-03 | Ground-truth verification: Claude Code extractions match leaked source at **85-95%** |
66+
| 2026-02 | Open-sourced **System Prompt Open** gallery with 45 extracted system prompts |
67+
| 2026-01 | Paper and **JustAsk** framework released. Initial extraction of 45 frontier LLMs |
4768

4869
## Related Work
4970

assets/favicons/ai21_labs.png

553 Bytes
Loading

assets/favicons/aion_labs.png

735 Bytes
Loading

assets/favicons/alibaba.png

496 Bytes
Loading

assets/favicons/allen_ai.png

509 Bytes
Loading

assets/favicons/amazon.png

1002 Bytes
Loading

assets/favicons/anthropic.png

1.3 KB
Loading

assets/favicons/arcee_ai.png

1.12 KB
Loading

0 commit comments

Comments
 (0)