Text Prompts

A curated library of 7 text prompt templates — system prompts, creative writing, technical docs, marketing copy, research synthesis, summarization, and a foundational prompt-engineering guide. Templates work across Claude, ChatGPT (GPT-5.x), Gemini, and open models like Llama — simply replace [PLACEHOLDERS] with your own content.

---

⚡ Quick Start

1. New to prompting? Start with the Prompt Engineering Guide.
2. Pick a template from the categories below.
3. Copy the prompt, replace the [PLACEHOLDERS], and run it.
4. Iterate — tighten the role, add examples, or strengthen the output-format contract based on results.

---

Platform Guide

Modern frontier models respond to the same core techniques; the differences are mostly in formatting conventions and context handling.

Claude (Anthropic)

• Responds especially well to XML-style tags for structure: <context>…</context>, <examples>…</examples>, <instructions>…</instructions>.
• Put long documents/context before the instructions for best results.
• Prefill the start of the assistant response (API) to lock output format.
• System prompts define role and rules; keep task-specific detail in the user message.

ChatGPT / GPT-5.x (OpenAI)

• Markdown headers and bullet lists work well as structure.
• Use the system/developer message for persistent role and rules.
• Explicit output-format instructions (“Respond only with valid JSON matching this schema…”) are honored reliably; structured outputs are available via the API.

Gemini (Google)

• Handles very long contexts well — you can include large documents inline.
• Prefers clear sectioned prompts with headers; state the persona, task, context, and format in that order.

Open models (Llama, etc.)

• Be more explicit and literal; avoid relying on implied conventions.
• Few-shot examples matter more — show, don’t just tell.
• Keep system prompts short and rule-like; long nuanced system prompts degrade more than on frontier models.

Universal principles

1. Role — tell the model who it is (“You are a senior technical writer…”).
2. Structure — separate context, instructions, examples, and format with clear delimiters.
3. Few-shot — show 1–3 input→output examples of exactly what you want.
4. Chain-of-thought — for complex reasoning, ask the model to work step by step before answering.
5. Output contract — specify the exact format, length, and constraints of the response.

The Prompt Engineering Guide covers each of these in depth with copyable patterns.

---

How to Use These Prompts

1. Find a template from the categories below.
2. Click through to the individual template doc.
3. Copy the prompt — most work as either a system prompt or a user message.
4. Replace placeholders — [TOPIC], [AUDIENCE], [PRODUCT], etc. — with your content.
5. Iterate — the Tips section in each doc lists the highest-leverage adjustments.

---

✍️ Text Prompt Templates

Prompt Engineering Guide

The five foundations of effective prompting — role, structure, few-shot examples, chain-of-though...

System Prompts

Reusable system-prompt patterns for assistants, domain experts, and structured-output services. A...

Creative Writing

Fiction, poetry, and narrative voice — prompts that treat the model as a collaborator with a brie...

Technical Documentation

READMEs, API references, tutorials, and architecture docs. The pattern throughout: give the model...

Marketing Copy

Landing pages, ads, emails, and product descriptions. Marketing prompts need three inputs to work...

Research Synthesis

Literature reviews, competitive analysis, and multi-source synthesis with explicit evidence gradi...

Summarization & Transformation

Summaries, rewrites, tone shifts, and format conversions — the workhorse tasks. The key controls:...

---

Contribute! Refined one of these templates or built a great new one? Share it via PR!