The Language of Simplicity

These prompts were written for this platform, so the structure is already right - the `### TASK ###` style headings read cleanly here. What to add is where the material comes from and where the output goes: a sales report sitting in Drive does not need pasting, and an email for the sales team can be drafted straight into Gmail. The simplicity gain is real, because pointing at a file is shorter and less error-prone than pasting one.

• Name the exact file, tab and range - not "my report".
• For internal figures, say "do not search" explicitly.
• Draft emails into Gmail as drafts, never sent.
• Upload the document rather than pasting it.
• Save recurring output formats as a Gem.

### TASK ###

Analyse the Q3 sales report. I am attaching it - do not ask me to paste it. If you are reading it from Drive instead, the file is "Q3 Sales Report" and the figures are on the "Summary" tab; tell me if you open anything else.

### REQUIRED OUTPUT ###

1. **Net profit** - the final figure, exactly as written, with the line item or page it came from.
2. **Top three products by sales** - bulleted, each with its figure. If revenue ranking and unit ranking differ, give both and say which you treated as primary.
3. **Overall conclusion** - one sentence.
4. **Discrepancies** - anywhere two parts of the document give different figures for the same thing. Quote both. This is the thing I most want found.

### DO NOT SEARCH ###

Use nothing but the attached document. Do not ground any figure in Google Search, do not add industry context, do not compare my numbers to anything external. Everything here must be traceable to a line in my file. A figure quietly corrected against a web source is the worst possible failure in this task, because it will look right.

### RULES ###

* Figures exactly as written, including lakh, crore or million. No conversions.
* Anything absent from the document: write "not stated in the report". Do not compute or estimate it.
* No commentary or recommendations.

### OUTPUT ###

Into a Google Doc called "Q3 analysis", with each extracted figure followed by its source location in the original. I am forwarding this and every number needs to be checkable.

Save this output format as a Gem called "Quarterly report extract" - I run this every quarter and the format should not change between them.

### TASK ###

A motivational email to the sales team for the start of Q4.

**Goal:** each person ends Q4 above target.
**Key message:** acknowledge their Q3 effort, then state plainly that Q4 carries the year.
**Call to action:** each team member books a 15-minute goal-setting meeting with their manager this week.
**Tone:** professional, energetic, encouraging - direct rather than effusive.
**Length:** under 200 words plus a subject line.

### NO IDIOMS - THIS IS THE MAIN INSTRUCTION ###

Banned: hit the ground running, knock it out of the park, give it 110%, move the needle, take it to the next level, all hands on deck, circle back, let's crush it.

These are the phrases I would have used myself, and they are the reason the email would be skimmed and forgotten. Write the literal thing underneath each one instead. "Hit the ground running" means start in week one rather than week three. "Give it 110%" means nothing I can actually ask for.

### USE MY ACTUAL DATA ###

Pull the Q3 numbers from the Sheet I am attaching - team total against target, and the two or three individual results worth naming. Use real figures rather than placeholders, but:

* Quote them exactly as they appear. Do not recalculate anything.
* Tell me which cells you used, so I can check them before this goes out.
* If naming individuals would be awkward because the spread is wide, tell me and leave them out. A motivational email that highlights three people and implicitly exposes the rest does the opposite of what I want.

### THEN ###

1. "Idiom check" - quote any phrase in your draft doing a cliché's job, including ones not on my list.
2. One line on how "Q4 is the most important quarter" reads in your version - encouragement or pressure - and an alternative if it tips wrong.

### OUTPUT ###

Save it as a draft in Gmail, addressed to the sales team group. Do not send it. I want to read it once with the Q3 numbers in front of me before anyone else does.

### TASK ###

A Python script that scrapes news headlines.

### SPECIFICATION ###

* **Libraries:** `requests` and `BeautifulSoup` (bs4). Not Selenium, not Scrapy.
* **Request:** HTTP GET to `https://timesofindia.indiatimes.com/`.
* **Extract:** the text of every `<h2>` element containing a headline.
* **Output:** print each headline on its own line.

### ROBUSTNESS ###

* A `User-Agent` header - a bare `requests` call is often refused.
* A timeout, and the status code checked before parsing.
* Handle connection errors, non-200 responses, and finding zero `<h2>` elements. The last is the most likely failure and the easiest to mistake for the site being down.
* Strip whitespace, skip empty strings.

### GROUND THIS PART - IT IS WHY I AM USING THIS PLATFORM ###

Use Google Search before writing the parsing code:

1. Check the current HTML structure of that homepage. Are the headlines actually in `<h2>` tags, and what class names do they carry? My specification above is a guess and I would rather have the real selector than a clean script that returns nothing.
2. Check the site's robots.txt and terms of use, and tell me plainly what they say about automated access. Cite them. Do not reassure me - if scraping it is disallowed, say so and tell me whether there is a feed or API I should be using instead.
3. If the page content is loaded by JavaScript rather than present in the initial HTML, say so clearly. In that case `requests` and BeautifulSoup will not work at all and the whole approach needs to change - that is worth knowing before I write anything.

If search cannot settle the selector, say so rather than guessing confidently.

### STRUCTURE THE SCRIPT FOR CHANGE ###

Put the tag and class in one named constant at the top, commented as the line to edit. Include a diagnostic block that prints the match count and the first two raw matches, so a broken selector shows itself immediately. Then explain in plain English how I find the selector myself with the browser inspect tool.

### OUTPUT ###

The script in the chat so I can iterate on it, and the search findings as a short separate section above it with their sources.