1,000 structured AI prompt templates across 75 categories, stored as Markdown and designed for real work.
PromptVault is an open-source, Markdown-first prompt knowledge base for research, software development, business strategy, HR, finance, marketing, analytics, regulated industries, professional operations, and personal productivity.
Instead of hiding prompts inside an application or database, PromptVault stores every prompt as an individual Markdown page. You can browse the library directly on GitHub, clone it for offline use, adapt templates in any text editor, or consume the machine-readable catalog in your own application.
Browse the prompt catalog · Read the user guide · View contribution guidelines · See release notes
Most prompt collections are long lists of short instructions with limited context. PromptVault is structured around practical execution. Templates may include:
- role and domain context;
- objective and decision to be supported;
- reusable
{{placeholders}}; - known facts and assumptions;
- scope, constraints, and exclusions;
- evidence and source requirements;
- analysis steps and decision criteria;
- risks, limitations, and counterarguments;
- required output structure;
- success measures and quality checks;
- follow-up actions and human-review gates.
The result is a library that works as both a ready-to-use collection and a reference for designing better prompts.
- 1,000 individual prompt templates
- 75 categories
- 1 Markdown page per prompt
- Model-agnostic templates
- Reusable placeholders using
{{placeholder_name}} - Central catalog for human browsing
- JSON catalog for applications and automation
- Research-grade patterns for evidence, sources, uncertainty, and contradiction handling
- Professional safeguards for high-impact and regulated use cases
- No runtime, API key, account, or database required
PromptVault spans broad functional and industry use cases, including:
Business strategy, entrepreneurship, leadership, project management, change management, operations, procurement, negotiation, decision making, communications, and executive reporting.
Software development, cloud and DevOps, APIs and integration, enterprise architecture, cybersecurity, privacy, quality engineering, automation, open source, GitHub, technical research, data science, and analytics.
Product management, product discovery, UX research, marketing, sales, customer service, market intelligence, customer discovery, retail, ecommerce, media, and creator workflows.
HR, HR systems, workforce planning, talent management, organization design, career development, education, knowledge management, and learning.
Finance, investing, banking, insurance, fintech, company intelligence, investment research, academic research, advanced research methods, economics, and social sciences.
Healthcare, life sciences, pharma, biotechnology, manufacturing, supply chain, energy, utilities, sustainability, agriculture, food, real estate, construction, automotive, mobility, aerospace, hospitality, travel, telecom, government, and public policy.
Open CATALOG.md for the complete category index.
promptvault/
├── prompts/
│ ├── README.md
│ ├── writing/
│ │ ├── README.md
│ │ ├── 01-blog-post-builder.md
│ │ └── ...
│ ├── software-development/
│ ├── deep-research/
│ ├── healthcare-life-sciences/
│ ├── enterprise-architecture/
│ └── ... 70 more categories
├── docs/
│ └── USER_GUIDE.md
├── CATALOG.md
├── prompts.json
├── GITHUB_METADATA.md
├── CONTRIBUTING.md
├── CHANGELOG.md
├── RELEASE_NOTES_v0.5.0.md
├── LICENSE
└── README.md
- Open CATALOG.md.
- Select a category.
- Open a prompt page.
- Copy the text inside the Prompt or Copy-ready prompt block.
- Replace every
{{placeholder}}with your context. - Paste the completed prompt into your preferred AI assistant.
- Select Code → Download ZIP on GitHub.
- Extract the archive.
- Open
CATALOG.mdorprompts/README.mdin a Markdown editor. - Search the repository using Visual Studio Code, Obsidian, Typora, or another Markdown-capable tool.
git clone https://github.com/<your-github-username>/promptvault.git
cd promptvaultReplace <your-github-username> after publishing the repository.
Search titles and content with ripgrep:
rg -n "market research|competitive analysis" prompts/Find placeholders used across the library:
rg -o '\{\{[a-zA-Z0-9_]+\}\}' prompts/ | sort | uniqList prompts in a category:
find prompts/software-development -name '*.md' -type f | sortA template may contain placeholders such as:
Analyze {{company}} in the {{industry}} sector for {{decision}}.
Use a {{time_horizon}} time horizon and evaluate it against
{{decision_criteria}}.
Replace the placeholders with specific information:
Analyze Microsoft in the enterprise software and cloud sector for
possible inclusion in a five-year long-term equity portfolio.
Use a five-year time horizon and evaluate it against revenue durability,
free cash flow, competitive position, capital allocation, valuation,
and downside risk.
Good inputs are specific enough to constrain the response while leaving the model room to perform the requested analysis.
For important tasks, use PromptVault as a workflow rather than a one-shot command.
Choose the prompt that most closely matches the real decision or deliverable, not merely the broad topic.
Replace every placeholder. Add audience, geography, date range, source limitations, constraints, and desired output format.
Attach or paste the relevant documents, data, notes, transcripts, code, policies, or research sources. Do not ask the model to infer information you can provide directly.
Run the completed prompt in your chosen AI assistant.
Check factual accuracy, source traceability, calculations, assumptions, omissions, risks, and instruction adherence.
Use focused follow-ups such as:
Challenge the three most important assumptions in the analysis.
Identify which conclusions would change if {{assumption}} were false.
Rewrite the recommendation for {{executive_audience}} in no more than
{{word_limit}} words.
Keep the customized prompt and validated output with the relevant project files. Remove confidential information before sharing publicly.
A strong prompt commonly contains the following components:
Role or responsibility
+ Context
+ Objective
+ Inputs
+ Constraints
+ Method
+ Evidence rules
+ Output contract
+ Quality criteria
+ Failure controls
Example:
You are supporting {{team}} with {{decision}}.
Context:
{{context}}
Use the following sources:
{{sources}}
Evaluate these alternatives:
{{alternatives}}
Apply these criteria and weights:
{{criteria_and_weights}}
Return:
1. Executive summary
2. Evidence table
3. Options comparison
4. Risks and counterarguments
5. Recommendation
6. Open questions
Do not invent missing facts or sources. Mark unsupported claims as
Unverified and explain how they can be validated.
prompts.json provides a machine-readable catalog for search tools, websites, desktop applications, agents, and automation workflows.
A catalog entry follows this general shape:
{
"id": "category-01",
"title": "Prompt title",
"category": "Category name",
"path": "prompts/category/01-prompt-title.md",
"placeholders": [
"context",
"objective",
"output_format"
]
}import json
from pathlib import Path
catalog = json.loads(Path("prompts.json").read_text(encoding="utf-8"))
research_prompts = [
item for item in catalog
if "research" in item["category"].lower()
]
for item in research_prompts[:10]:
print(item["title"], "->", item["path"])import catalog from "./prompts.json" with { type: "json" };
const securityPrompts = catalog.filter((item) =>
item.category.toLowerCase().includes("security")
);
console.log(securityPrompts);Research-oriented prompts should be adapted to specify:
- the exact research question;
- decision being supported;
- geography and jurisdiction;
- date range and recency threshold;
- source hierarchy;
- acceptable and excluded sources;
- citation format;
- method of triangulation;
- treatment of contradicting sources;
- confidence or evidence-strength method;
- known limitations and unresolved questions.
Recommended source hierarchy:
- Laws, regulations, standards, official filings, audited reports, government datasets, original research, contracts, and official product documentation.
- Peer-reviewed research, respected professional bodies, established research institutions, and reputable analyst material.
- Established news organizations and expert practitioner publications.
- Vendor claims, community posts, aggregators, anonymous content, and AI-generated summaries, used primarily for discovery and verified independently.
Never treat a citation-shaped answer as verified evidence. Open the source, confirm that it exists, check the relevant passage, and verify that it supports the claim.
PromptVault includes templates touching healthcare, finance, legal work, cybersecurity, public policy, engineering, HR, and other consequential domains.
These templates are intended for:
- research planning;
- evidence organization;
- drafting and summarization;
- scenario analysis;
- requirements gathering;
- issue spotting;
- decision support;
- preparing questions for qualified professionals.
They are not substitutes for licensed medical, legal, financial, engineering, safety, or regulatory judgment. Validate material outputs against authoritative sources and route consequential decisions to appropriately qualified reviewers.
Do not place secrets, credentials, personal data, medical records, privileged legal material, controlled information, or confidential company data into an AI service unless your organization has approved the service and the specific data use.
The prompts are intentionally model-agnostic. They can be adapted for tools such as Microsoft Copilot, ChatGPT, Claude, Gemini, locally hosted models, or application-specific agents.
Results will vary by model, enabled tools, context-window limits, access to external sources, and organization-level policies. A prompt requiring current research will only produce source-grounded results when the chosen tool has appropriate search or document access.
Contributions are welcome. New prompts should:
- solve a specific, credible use case;
- use descriptive titles and stable IDs;
- use lowercase kebab-case filenames;
- expose reusable inputs as
{{placeholders}}; - define the required output;
- separate facts, assumptions, and recommendations where relevant;
- include evidence and uncertainty controls for research tasks;
- include human-review guidance for high-impact use;
- avoid secrets, personal data, fabricated claims, and copyrighted source text;
- be added to the category index,
CATALOG.md, andprompts.json.
See CONTRIBUTING.md for the repository contribution process.
Potential future improvements include:
- richer metadata and task classification;
- beginner, intermediate, and advanced difficulty levels;
- prompt quality scores and automated linting;
- related-prompt links;
- example inputs and validated example outputs;
- prompt chains and multi-step workflows;
- YAML and JSON prompt schemas;
- automated catalog generation;
- a local PromptVault desktop or web interface;
- favorites, collections, version history, and import/export;
- community review and domain-expert validation.
Current release: v0.5.0 — 1,000 Prompt Library
See RELEASE_NOTES_v0.5.0.md and CHANGELOG.md.
PromptVault is released under the MIT License. See LICENSE.
PromptVault is maintained as part of the PeaBox product and open-source experimentation ecosystem.