# AndreSzymanski15/ai-jailbreaks

> A curated collection of 35 paste-ready jailbreak prompts, system prompts, and AGENTS.md specs for 25+ models, built for AI security research and red-team testing.

- Magnitude: 4.4 out of 10 — Early signal
- Stars: 198 total · +28 stars measured 2026-10-19, 18:27–20:46 UTC, ≈ 32 by evening
- Star trust: star growth looks organic
- Category: Security · Created: 2026-10-19 · Last push: 2026-10-19
- GitHub: https://github.com/AndreSzymanski15/ai-jailbreaks · Page: https://gitnova.dev/en/r/AndreSzymanski15/ai-jailbreaks

## Useful for

- Test a model's resistance to prompt injection using ready prompts
- Assemble an AGENTS.md to probe coding-agent behavior
- Run an Ollama Modelfile with a custom SYSTEM block

## Why it’s here

- Star-counter measurements on 2026-10-19 (UTC), 18:27–20:46: 170 → 198 stars (+28). This is the change over that interval.
- Estimated end-of-day forecast: about +32 stars, using observed gains and the previous day.
- The repository is 0 days old and already has 198 stars. With less than two weeks of history, there's no usual pace to compare the spike against yet.
- Top new repositories this week: #128.
- About 112 forks a day — people are taking the code.

## Star trust

Star growth looks organic.

Star-trust labels are heuristics based on the repository’s behavior, not a check of every stargazer.

## Numbers

- Forks: 30
- Issues and pull requests: 0
- Watchers: 18
- Average over the last week: 32 per day
- Usual pace: too little history (under two weeks)
- Stars in the last hour (measured): 11

## Stars per day, last 1 day (oldest → newest, today is partial)

2026-10-19 … 2026-10-19: 28

## Spotted in now

- Top new repositories this week: #128

## More in this category

1. **YuJunZhiXue/dsh-purge** — 7.3 · Breakout · Security · JavaScript · +1,543 stars measured 2026-10-19, 00:06–20:43 UTC, ≈ 1,744 by evening
   A local DeepSeek Harness plugin that reversibly patches installed @deepseek-ai packages to strip built-in refusals and swap prompts per model. Intended for red-team research and robustness evaluation in an authorized local environment.
   Full card: https://gitnova.dev/en/r/YuJunZhiXue/dsh-purge.md
2. **interplains/zero-breach** — 6.3 · Early signal · Security · Python · +266 stars measured 2026-10-19, 00:06–20:43 UTC, ≈ 303 by evening
   A local, read-only Python scanner that looks for traces of the AdFection / Candiru spyware campaign on a machine: browser history and cookies, caches, autostart entries, file hashes and the DNS resolver cache. No admin rights, no…
   Full card: https://gitnova.dev/en/r/interplains/zero-breach.md
3. **sakidekin/CVE-2026-41940-PoC** — 5.7 · Early signal · Security · Python · +53 stars measured 2026-10-19, 00:05–20:44 UTC, ≈ 62 by evening
   PoC exploit for CVE-2026-41940, an authentication bypass in cPanel & WHM (CVSS 10.0) that grants unauthenticated root-level WHM access via CRLF injection into server-side sessions. The tool scans targets and performs post-exploitation…
   Full card: https://gitnova.dev/en/r/sakidekin/CVE-2026-41940-PoC.md
4. **gichirimaku/wp2shell-PoC** — 5.7 · Early signal · Security · Python · +93 stars measured 2026-10-19, 00:05–20:44 UTC, ≈ 108 by evening
   Proof-of-concept exploit for the wp2shell vulnerability chain in WordPress Core (CVE-2026-63030 and CVE-2026-60137), leading from a pre-auth SQL injection to RCE. Intended for testing your own or authorized sites.
   Full card: https://gitnova.dev/en/r/gichirimaku/wp2shell-PoC.md
5. **cloudflare/security-audit-skill** — 5.0 · Breakout · Security · JavaScript · +1,186 stars measured 2026-10-19, 00:07–20:44 UTC, ≈ 1,343 by evening
   A coding-agent skill that turns the agent into a security auditor, running a multi-phase code audit with independently verified, machine-readable findings.
   Full card: https://gitnova.dev/en/r/cloudflare/security-audit-skill.md

---

Magnitude (0–10) measures how fast and how unusually interest in a repository is growing right now. It is not a quality score.
Days are UTC. “So far today” is a fact; “expected by the end of the day” is a forecast.
Summaries and use cases are written by an LLM (DeepSeek V4.1 Flash) from the README and may be inaccurate: verify specific claims (benchmarks, speed, hardware) in the repository itself.
Data as of 2026-10-19 20:58 UTC, updated every 30 minutes.