LLM Prompts Collection

A working archive of prompts for large language models. Sorted into eleven categories, kept as plain text so what you copy is exactly what was written.

English · Türkçe · Español · 简体中文 · Русский

Index · Categories · How to use · Provenance · Contributing · Disclaimer

---

✨ Highlights

• Plain text, on purpose - every prompt is a .txt file. GitHub renders Markdown, and a prompt containing ** or # would be copied out mangled. What you see here is byte for byte what goes into the model.
• Sorted by what you are trying to do, not by which model you use. Most of these work across GPT, Claude, Gemini and open models with no changes.
• Versioned by hand - files carry V1, V2, V3 suffixes where a prompt was reworked, so you can see how one evolved and go back to an earlier take.
• Nothing to install - clone it, or read any file straight on GitHub and copy it.

---

📂 Categories

Every file in the archive with a one-line description of what it does and how long it is: INDEX.md.

Folder	What is in it
!Jailbreak-Prompts	Alignment and boundary testing prompts, kept for research. Read the disclaimer before using any of these.
Writing-and-Marketing	Copy, SEO, social media, review writing, humanising output
Interpersonal-Relations	Social strategy and analysis prompts
Prompt-Engineering	Prompts that write or refine other prompts
Persona-Mimicry	Voice and persona construction
Research-and-Analysis	Academic research and structured analysis
Career-and-Business	CV rewriting, review and hiring-side prompts
Programming	Architecture and code-focused prompts
Finance-and-Investment	Market and momentum analysis
Gaming-and-FRP	Game mastering and roleplay
Token-Reduction	Prompts that cut output length and cost

The jailbreak folder is prefixed with ! so it sorts to the top of the listing. That character is legal on Windows, macOS and Linux, unlike the * that was there before.

---

⚡ How to use

1. Open the folder that matches what you are trying to do.
2. Open the file. GitHub shows .txt raw, with line numbers, so nothing is reformatted.
3. Copy the whole file and paste it into your model.
4. Where several versions exist, the highest number is the most recent. The older ones are kept because a simpler version is sometimes the better one.

git clone https://github.com/Miabeyefendi/LLM-Prompt-Collection.git

---

🧾 Provenance

This is a curated archive, not a body of original work. Some of these prompts I wrote. Others I collected from around the internet over time, and in most cases the original author is not recorded.

What that means in practice:

• The collection, its structure and the files I wrote are mine, and are licensed as stated below.
• A prompt written by someone else remains theirs. Collecting a text into an archive does not transfer its copyright, and nothing here should be read as a claim over work I did not write.
• If you recognise something here as yours and want it credited or removed, open an issue and I will do it.

---

🤝 Contributing

Contributions are welcome. Read CONTRIBUTING.md and CODE_OF_CONDUCT.md first.

1. Put the prompt in the folder that matches its purpose.
2. Save it as .txt, named for what it does, with a V1 suffix.
3. If you did not write it, say so in the pull request and name the source if you know it.

Do not add a prompt whose only purpose is to produce content that is illegal, that targets a specific real person, or that is designed to cause harm. That is not a boundary-testing prompt, it is just the harm with extra steps.

---

🛡️ Security

Found something that should not be here, or a file that leaks personal data? Do not open a public issue. Follow the private process in SECURITY.md.

---

⚠️ Disclaimer

Read this before using anything in the jailbreak folder.

This archive exists for research and education. The alignment testing prompts are here so that people studying model safety can see what these attempts actually look like, which is difficult to do from papers alone.

• I do not endorse using any of this to produce illegal, harmful, threatening or deceptive content.
• Every model provider has terms of service. Using these prompts may breach them, and that is between you and the provider.
• You are responsible for what you generate. Not the model, not this repository, and not me.
• These are alignment tests, not exploits against people. Prompts aimed at a specific real person, or at producing something that causes concrete harm, do not belong here and will be removed.

This material is provided "as is", without warranty of any kind, to the maximum extent permitted by applicable law.

---

📜 License

Everything I wrote here, the prompts that are mine along with the categories, the index and the shape of the archive, is released under CC0 1.0. Public domain. Nothing to ask permission for, no attribution to keep, no conditions.

CC0 cannot cover two things:

• Prompts I collected rather than wrote. Those belong to whoever wrote them. Putting a text into an archive does not give me the standing to waive someone else's rights over it. See Provenance.
• What the prompts produce. Whatever a model writes for you is between you and whoever runs that model.

Credit is always welcome and never required.

---

📬 Contact

• GitHub: @Miabeyefendi

_{Collected by Miabeyefendi}