Best prompts for ChatGPT using few-shot example templates

Task: Rewrite the following email into a concise, professional message. Keep important details, remove filler, and provide an optional subject line. Example 1 Input: "hey team, sorry for the late reply — ive been swamped. can we push the meeting to next week? i have conflicts on wed and thursday but any other day is fine. also need the slide deck updated with the new figures if possible. thanks" Output: Subject: Request to Reschedule Meeting Hi team, Apologies for the late reply. Could we reschedule the meeting for next week? I’m unavailable Wednesday and Thursday but available any other day. Please also update the slide deck with the new figures. Thanks, <Name> Example 2 Input: "the report i sent had some mistakes—i forgot to include q4 numbers. i attached an updated file. please use that version for review. sorry for the confusion." Output: Subject: Updated Report with Q4 Numbers Hi, I sent an earlier report that omitted Q4 numbers. I’ve attached the corrected file—please use it for your review. Apologies for the confusion. Best, <Name> Template Input: "<paste original email here>" Output: Subject: <optional short subject> <professional, concise rewrite preserving key facts> Now rewrite the following email:

Task: Given a short Python function with a bug, return: 1) Corrected code only (delimited by triple backticks), 2) One-paragraph explanation of the bug and fix, 3) A simple assert-based test showing expected behavior. Example 1 Input: def add_items(lst, item): for i in item: lst.append(i) return lst Output: ```python def add_items(lst, item): if isinstance(item, (list, tuple, set)): for i in item: lst.append(i) else: lst.append(item) return lst ``` Explanation: The original code iterated over 'item' assuming it was iterable; if 'item' is a scalar (e.g., int or str) this misbehaves. The fix handles both iterables and single values. Test: assert add_items([1,2], 3) == [1,2,3] assert add_items([1], [2,3]) == [1,2,3] Example 2 Input: def mean(values): return sum(values) / len(values) Output: ```python def mean(values): values = list(values) if len(values) == 0: raise ValueError('mean() arg is an empty sequence') return sum(values) / len(values) ``` Explanation: The original fails on empty lists or non-reusable iterators. The fix converts to list and raises a clear error for empty input. Test: try: mean([]) except ValueError: pass Template Input: <insert buggy python function here> Now fix:

Task: Parse the provided receipt text and output a JSON object with these fields: date (YYYY-MM-DD), vendor, currency, items (array of {name, qty, unit_price, total}), subtotal, tax, total. If a field is missing, set null. Example 1 Input: "Starbucks 07/14/2024 1x Latte $4.50 2x Cookie $3.00 Subtotal $7.50 Tax $0.68 Total $8.18" Output: { "date": "2024-07-14", "vendor": "Starbucks", "currency": "USD", "items": [ {"name":"Latte","qty":1,"unit_price":4.50,"total":4.50}, {"name":"Cookie","qty":2,"unit_price":1.50,"total":3.00} ], "subtotal": 7.50, "tax": 0.68, "total": 8.18 } Example 2 Input: "IKEA 2024-05-02 Receipt #12345 Chair x2 49.99each 99.98 VAT 19% Total 118.98" Output: { "date": "2024-05-02", "vendor": "IKEA", "currency": "USD", "items": [{"name":"Chair","qty":2,"unit_price":49.99,"total":99.98}], "subtotal": 99.98, "tax": 19.00, "total": 118.98 } Template Input: "<paste receipt text>" Now parse:

Task: Translate the Input sentence to <language> using the <tone> tone and <locale>. Keep meaning identical and adapt idioms. Example 1 Input: "Can you send the report by Friday?" Language: Spanish Tone: Formal Locale: es-ES Output: "¿Podría enviar el informe antes del viernes, por favor?" Example 2 Input: "Hey, are we still on for tomorrow?" Language: Spanish Tone: Informal Locale: es-MX Output: "¿Todavía nos vemos mañana?" Template Input: Text: "<paste text>" Language: <language> Tone: <formal|informal> Locale: <locale code> Now translate:

Task: Translate the requirement into SQL. If creating tables, include types, primary keys, indexes, and NOT NULL where appropriate. If selecting, return the query and a short comment explaining joins/filters. Example 1 Input: "Create a users table with id, email, name, signup_date, and is_active" Output: -- Assumptions: emails unique; id is serial primary key CREATE TABLE users ( id SERIAL PRIMARY KEY, email VARCHAR(255) NOT NULL UNIQUE, name VARCHAR(255) NOT NULL, signup_date DATE NOT NULL, is_active BOOLEAN NOT NULL DEFAULT TRUE ); Example 2 Input: "Find customers who purchased product_id 42 in last 30 days with total spent per customer" Output: -- Assumptions: tables orders(id, customer_id, order_date), order_items(order_id, product_id, price, qty) SELECT c.id AS customer_id, c.name, SUM(oi.price * oi.qty) AS total_spent FROM customers c JOIN orders o ON o.customer_id = c.id JOIN order_items oi ON oi.order_id = o.id WHERE oi.product_id = 42 AND o.order_date >= CURRENT_DATE - INTERVAL '30 days' GROUP BY c.id, c.name; Template Input: "<plain-English requirement>" Generate SQL:

Task: Read the transcript and output JSON with two arrays: actions [{owner, action, due}] and decisions [string]. Keep actions specific and assignable. Example 1 Input: "Anna: We should update the onboarding doc. Mike: I'll handle the first draft by next Tuesday. Sarah: Let's remove the deprecated steps." Output: { "actions": [ {"owner":"Mike","action":"Draft updated onboarding document","due":"next Tuesday"}, {"owner":"Sarah","action":"Identify deprecated steps to remove","due":null} ], "decisions": ["Update onboarding document to remove deprecated steps"] } Example 2 Input: "PM: Q3 campaign goes live Sept 1. Designer: I'll provide assets by Aug 20. Ops: Need budget sign-off by Aug 5." Output: { "actions": [ {"owner":"Designer","action":"Provide campaign assets","due":"2025-08-20"}, {"owner":"Ops","action":"Obtain budget sign-off","due":"2025-08-05"} ], "decisions": ["Q3 campaign launch date set to 2025-09-01"] } Template Input: "<paste transcript>" Now extract: