An open-source prompt database — browse on GitHub or fetch any prompt programmatically via the free JSON API.
No account. No key. No cost.

⭐ If you found this database helpful, please leave a star on the GitHub repository! ⭐

Quick Start  ·  How to Use  ·  Browse Prompts  ·  JSON API  ·  Contributing

---

Quick Start

No installation. No account. Just copy and use.

1. Find a category in the table below
2. Open any .md file — the prompt lives inside the blockquote
3. Paste it into ChatGPT, Claude, Gemini, or any LLM
4. Tweak and go

Developer? Skip the browsing — use the free JSON API to fetch prompts programmatically with a single fetch() call.

---

How to Use

No account. No API key. No setup required.

Browse on GitHub

Click a category in the table below → open any .md file → copy the blockquote text → paste into your LLM of choice.

Understand the variable syntax

Prompts use ${variable} placeholders so you can fill in specifics:

Syntax	Meaning
${topic}	Required input — replace with your value
${audience:developers}	Optional — default is developers, override as needed

Example: Write a blog post about ${topic} for ${audience:developers}. Replace ${topic} with your subject. Keep or change the ${audience} default.

Search locally

# Full-text search across all prompt files
grep -ri "keyword" prompts/

# Search the machine-readable registry (one JSON record per line)
grep "keyword" data/registry.jsonl

Use the data exports

Goal	File to use
Import into Excel or Google Sheets	data/prompts.csv
Load in Python / pandas	data/prompts.csv or data/prompts.json
Build a search tool or internal API	data/registry.jsonl — stable IDs, deduplicated
Filter by category, variables, or status	data/registry.jsonl — filter on category, variables, lifecycle_status

---

Practical Examples

These examples show the structure of a useful prompt: clear inputs, explicit constraints, a defined output format, and honest handling of missing evidence.

Code review with actionable findings

You are a senior code reviewer.

Review this change:
- Diff: ${diff}
- Project context: ${context}
- Risk priorities: ${risk_priorities:correctness,security,maintainability}

Rules:
1. Identify only issues supported by the diff or project context.
2. Do not invent files, tests, APIs, or runtime behavior.
3. Separate blocking issues from suggestions.
4. If evidence is insufficient, say so explicitly.

Return Markdown with exactly these sections:
## Summary
## Blocking issues
## Non-blocking suggestions
## Missing evidence

For every issue, include: severity, file and line when available, why it matters, and a concrete fix.

Reliable document-to-JSON extraction

Extract the requested facts from ${document}.

Return one valid JSON object matching this shape:
{
  "entities": [{"name": "string", "type": "string", "value": "string"}],
  "dates": [{"label": "string", "value": "YYYY-MM-DD or null"}],
  "uncertain_items": ["string"]
}

Use null or an empty array when the document does not provide a value. Do not infer facts from general knowledge. Preserve the document's wording in `uncertain_items`. Return JSON only, with no Markdown fences or commentary.

Evidence-first research brief

Prepare a concise research brief about ${topic} for ${audience:technical decision-makers}.

Use the supplied sources: ${sources}. For each important claim, include its source URL. Separate:
- verified facts directly supported by a source;
- reasonable interpretations;
- open questions or missing evidence.

Do not present an estimate, opinion, or single-source claim as an established fact. If the sources are insufficient, state what cannot be concluded. End with three decision-relevant questions, not a generic conclusion.

Translation that preserves structure

Translate ${text} from ${source_language:English} to ${target_language:French}.

Preserve exactly:
- Markdown headings, lists, links, code blocks, and HTML tags;
- variables such as ${name} and `${format:json}`;
- product names, file paths, and code identifiers.

Do not translate code or alter placeholder names. If a phrase has two materially different translations, choose the most natural one for the target locale and add one short note after the translated text. Otherwise, return only the translation.

---

Categories

2,103 curated prompts across 19 categories — updated regularly.

Category	Prompts	Browse
Coding & Development	857	→ prompts/coding-development
Image & Design	478	→ prompts/image-design
Writing & Content	253	→ prompts/writing-content
Data & Analytics	224	→ prompts/data-analytics
Marketing & Social	191	→ prompts/marketing-social
General	159	→ prompts/general
AI & Automation	147	→ prompts/ai-automation
Business & Career	140	→ prompts/business-career
Documentation	137	→ prompts/documentation
Security	128	→ prompts/security
Health & Wellness	115	→ prompts/health-wellness
Research & Analysis	109	→ prompts/research-analysis
Sales & Business	106	→ prompts/sales-business
Games & Fun	100	→ prompts/games-fun
Product & Strategy	95	→ prompts/product-strategy
Travel & Places	94	→ prompts/travel-places
Philosophy & Humanities	89	→ prompts/philosophy-humanities
Food & Recipes	88	→ prompts/food-recipes
Education & Learning	70	→ prompts/education-learning

Data Exports

The canonical registry is the deduplicated, machine-readable source of truth. The original corpus files are preserved for compatibility.

File	Format	Records	Use case
data/registry.jsonl	JSONL	2,103	Stable IDs, deduplication, variables, provenance
data/prompts.json	JSON array	3,470	Original imported corpus
data/prompts.csv	CSV UTF-8	3,470	Excel, pandas, Sheets, SQL imports
data/sources.json	JSON array	5	Source inventory and review status

Canonical registry record

{
  "id": "opl_0123456789ab",
  "slug": "prompt-title",
  "title": "Prompt title",
  "prompt": "Full prompt text",
  "category": "Category name",
  "folder": "category-folder-name",
  "source_ids": ["src_0123456789ab"],
  "provenance_status": "needs-review",
  "lifecycle_status": "draft",
  "revision": 1
}

---

Free JSON API

Every prompt is available as a structured JSON endpoint — served by GitHub Pages, no backend, no API key, no rate limits.

Base URL: https://belyagoubiabdelilah.github.io/open-prompt-library

Endpoints

Endpoint	Description
/v1/stats.json	Global counts, category list, all endpoint URLs
/v1/categories/index.json	All 19 categories with prompt counts
/v1/categories/{folder}.json	All prompts in a category (full text)
/v1/index.json	Paginated prompt list — page 1, 100 per page
/v1/index/page-N.json	Subsequent pages (follow next link)
/v1/prompts/{id}.json	Single prompt by stable ID (full text)

Quick start for developers

// Fetch all coding prompts
const res = await fetch(
  'https://belyagoubiabdelilah.github.io/open-prompt-library/v1/categories/coding-development.json'
);
const { prompts } = await res.json();
console.log(prompts[0].prompt); // full prompt text, ready to use

import requests

# Discover all available categories
data = requests.get(
    'https://belyagoubiabdelilah.github.io/open-prompt-library/v1/categories/index.json'
).json()

for cat in data['categories']:
    print(cat['name'], cat['prompt_count'], cat['url'])

# Fetch a single prompt by ID
curl https://belyagoubiabdelilah.github.io/open-prompt-library/v1/prompts/opl_8fbf9edcd378.json

Prompt record shape

{
  "id":                "opl_0123456789ab",
  "slug":              "prompt-title",
  "title":             "Prompt title",
  "prompt":            "Full prompt text, ready to paste into any LLM.",
  "category":          "Coding & Development",
  "folder":            "coding-development",
  "variables":         [{ "name": "topic", "default": null }],
  "provenance_status": "needs-review",
  "lifecycle_status":  "draft",
  "source_ids":        ["src_0123456789ab"],
  "revision":          1,
  "url":               "https://belyagoubiabdelilah.github.io/open-prompt-library/v1/prompts/opl_0123456789ab.json"
}

CORS — GitHub Pages serves all files with open CORS headers. Calls from any browser or origin work without a proxy.

Updates — The API rebuilds automatically on every push to main. The last_updated field in each response tells you exactly when the snapshot was taken.

---

Building Locally

No dependencies required — only Node.js.

git clone https://github.com/BELYAGOUBIABDELILAH/open-prompt-library.git
cd open-prompt-library
node scripts/build-registry.js
node scripts/validate-registry.js
node scripts/generate-tree.js
node scripts/build-api.js      # generates api/v1/ for local testing

The build is dependency-free and idempotent. build-registry.js creates data/registry.jsonl, data/sources.json, data/quarantine.jsonl, and data/registry-stats.json without modifying the source records.

---

Contributing

See CONTRIBUTING.md for the full guide. New submissions should include a source, license, intended use, variables, and a short example of expected output.

Single prompt	Open a PR directly
Batch of prompts	Open an issue first to coordinate
Found a duplicate or bug	Open an issue

All contributions are reviewed and deduplicated before merging.

---

License

MIT · Prompt sources are credited inline in each file.

---

Back to top

 ·   · 

Maintained by BELYAGOUBIABDELILAH  ·  Open source  ·  Always free

_{2,103 curated prompts · Free JSON API · No sign-up required}