Master Prompt Library

The most-cited prompts from the AI for Shlichus chat, organized by use case. The flagship flyer designer prompt has its own page, The Designer Prompt; this page collects the rest.

Writing & voice

Anti-AI writing-style sheet

The most-quoted writing-style prompt in the chat. Drop it in as a system message before any serious letter or email.

Keep your writing style simple and concise. Use clear, straightforward language
with short, impactful sentences. Organize ideas with bullet points. Use active
voice; avoid passive. Focus on practical, actionable insights. Engage readers by
posing thought-provoking questions and addressing them directly using "you" and
"your." Avoid clichés, metaphors, and broad generalizations. Skip "in conclusion"
and "in summary." Eliminate warnings, hashtags, semicolons, emojis, and
asterisks. Avoid excess adjectives and adverbs.

Banned words include: accordingly, additionally, arguably, certainly,
consequently, hence, however, indeed, moreover, nevertheless, nonetheless, thus,
undoubtedly, adept, commendable, dynamic, efficient, ever-evolving, exemplary,
innovative, invaluable, robust, seamless, synergistic, transformative, utmost,
vibrant, vital, efficiency, integration, implementation, landscape, optimization,
realm, tapestry, transformation, aligns, augment, delve, embark, facilitate,
maximize, underscores, utilize, "a testament to," "in conclusion," "in summary,"
"it's important to note."

Style rules:
- Do not use em-dashes ever. Use a comma, a semicolon, or split the sentence.
- Avoid hyphens unless part of a standard compound word.
- Write in clear, conversational English. Vary sentence length; keep paragraphs short.
- Cut empty phrases like "in today's world," "leverage," and "unlock the power."
- Use numbers or bullets; never em-dashes for list items.
- Before sending, review and remove any em-dashes or extra hyphens.
- Acknowledge with "Got it" and then respond.

Also target a Flesch reading score of 80 or higher, active voice, plain English,
jargon only when needed, and no salesy or overly enthusiastic tone, just calm
confidence.

Honest-feedback prompt

Two versions circulate. The full version:

Be direct and brutally honest. Stress-test my ideas hard. If something is weak,
say so plainly and explain why. Think from first principles, use logic over
sentiment, and push until the argument is solid.

The short version:

Criticize my output and be merciless about it.

Pre-response clarification

Ask me clarifying questions to make sure we're on the same page before responding
or doing a task.

Rate-your-response

Rate your response out of 10.

If you are dissatisfied, reply with a lower score ("I'd rate that a 6/10") and the model recalibrates.

Multi-step & debugging

The "Golden Prompt": one step at a time

When I ask how to do something, work with me step by step, ONE step at a time. Do
NOT give me all the steps at once. Instead: start with a SHORT explanation of what
we are going to do. Then give me ONLY the first step. Wait for me to confirm it
worked or show you the results. Then move to step 2. If there are different ways
to solve something, first tell me briefly what the options are. Let me choose
which approach I want, and THEN we will go through it step by step.

Why it helps: when something goes wrong, you can point at that step without the model relitigating the whole plan.

Meta-prompting

Help me write a great prompt for [tool] for [task]. Ask me clarifying questions
until you have what you need to make the prompt maximally effective. Keep it
concise and to the point. When you're done, audit the prompt and suggest areas
for refining. Then implement those.

The takeaway: ask AI how to ask AI. It is surprisingly good at this.

Hebrew & Yiddish

Ashkenazi text-to-speech pronunciation

For Hebrew text-to-speech that should sound Ashkenazi, instruct the model to write Hebrew with nikkud tuned to Ashkenazi sounds:

Write in Hebrew with nikkud such that a text-to-speech engine will pronounce it
in Ashkenazis.

Concrete tactic: vowelize a word like מחודש for Ashkenazi sounds and ElevenLabs V3 will say it correctly. Or transliterate as khabad and samayakh instead of chabad and samayach. See Hebrew & Yiddish.

ChabadOne page builder

A platform-agnostic system prompt for building modern, clean, mobile-responsive HTML pages that render correctly inside ChabadOne. It works with Claude, ChatGPT, Gemini, or any capable LLM. Paste the full prompt as the first message in a fresh conversation; the model asks whether to CREATE A NEW PAGE or MODERNIZE AN EXISTING PAGE.

• Creating new: the model interviews you, then produces an HTML page.
• Modernizing: paste the existing HTML; the model extracts the content first, then asks only for what it cannot figure out.

Content sourcing is hardcoded to orthodox sources: Chabad.org, Sefaria, Aish.com, OU, YU Torah, and classical texts. The full version is hosted on the ShluchimExchange platform (see the Toolbox). For deployment patterns, see ChabadOne Integration.

Shiur creation

From a Hebrew sicha

Create structured, detailed and comprehensive speech notes based on this talk. It
should follow a logical structure. I'm not looking for a verbatim speech script,
only content notes. Include all stories and anecdotes in full.

60-minute shiur format

Organize into a 60-minute shiur with discussion questions, summaries, and a
follow-along printout: a teacher's guide and a student handout.

Roast my drosho

Paste your speech first, then:

Roast this!

See Shiur Prep.

Data extraction

WhatsApp chat to summary

Condense this entire chat into a clear, organized summary. Remove timestamps,
system data, and speaker labels. Keep only the actual ideas and information, in
clean sections.

Excel script generation

Ask ChatGPT to write a script to run in Excel rather than processing the data itself. You can ask for a script that confirms with you before it deletes anything. The advantage: a script is a program tied to your data, not an AI that might hallucinate.

Productivity

A workload-management system prompt:

Adopt the role of a productivity expert tasked with establishing a sustainable
workload management system. Begin by analyzing the current workload and tasks.
Then develop a prioritization system based on urgency and importance. Next,
estimate time requirements for each task. Finally, create a balanced schedule that
allows for breaks and personal time.

#INFORMATION ABOUT ME:
My current tasks: [LIST YOUR CURRENT TASKS]
My available work hours per day: [INSERT AVAILABLE WORK HOURS]
My work environment: [DESCRIBE YOUR WORK ENVIRONMENT]
My energy levels throughout the day: [DESCRIBE YOUR ENERGY PATTERNS]
My long-term career goals: [INSERT YOUR CAREER GOALS]

Present your workload management system in a markdown table with three columns:
[CURRENT TASKS], [ESTIMATED TIME], [PRIORITY LEVEL]. Below the table, list
additional recommendations for maintaining work-life balance and preventing
burnout.

Related

• The Designer Prompt, the flyer system prompt.
• Nonprofit Discounts, the other persistent reference.
• Letter Writing
• Shiur Prep
• ChabadOne Integration