I Tested Over 100 Claude Code Prompts — These 6 Actually Made Me Money

Last month, I spent 78 hours testing every code-related skill Claude 3.5 claims to have. I wanted to know what actually works for solopreneurs trying to automate their way to more profit. Out of 127 prompts, only 6 delivered results I could plug into real projects and start earning from. The rest? Just noise.

I run three small SaaS tools that rely on lightweight automation. I don’t have a dev team. I need AI that writes reliable Python, scrapes data cleanly, and connects APIs without breaking every Tuesday. After testing dozens of workflows — from auto-generating Shopify meta tags to scraping local service leads — I found six code skills that consistently saved me time and created direct revenue.

1. Auto-Scrape Local Service Leads (and Sell Them)

I built a lead scraper that pulls HVAC, plumbing, and roofing contractor info from Google Maps using Claude-generated Python and Playwright. The prompt asked for "a headless browser script to extract business name, phone, address, and website from Google Maps search results for 'emergency plumber Austin'."

Claude returned working code in 3 tries. I ran it daily, cleaned the data with a second script (also Claude-built), and sold curated lists on MicroAcquire and LinkedIn for $197 per batch. First month: $1,182 from 6 sales. Cost? $0.48 in API fees and 1.2 hours of my time.

Pro tip: Use Playwright over Selenium. It’s faster and less likely to trigger anti-bot flags. And always add random delays between requests — Claude can generate that logic too.

2. Turn CSVs into Live Dashboards with Streamlit

I tested 14 dashboard tools. Streamlit, paired with Claude’s code, was the only one I could deploy solo in under 2 hours.

I had a messy CSV of customer support tickets. I asked Claude: "Write a Streamlit app that loads a CSV, shows ticket volume by day, agent response time, and a filter by status." It gave me a full script with st.line_chart, st.selectbox, and Pandas date parsing — all functional on first run.

I now charge $299 to build custom dashboards for small agencies using this same template. Took me two weeks to land my first client through a post on Indie Hackers.

3. Auto-Generate SEO Content from Product Feeds

I used to pay $0.10/word for Shopify product descriptions. Then I built a script that pulls product data from a CSV, uses Claude to generate unique descriptions, and pushes them back to the store via the Shopify Admin API.

Claude wrote the entire workflow: reading the CSV, batching products, calling the Anthropic API with a tight prompt ("Write a 60-word product description for {product} focused on durability and fast shipping"), and updating each item.

For a client with 1,200 products, I saved them $720 in copywriting fees. I charged $450 for the script and one-time setup. ROI was immediate.

4. Fix Broken JSON in Real Time

Half the API responses I deal with are malformed. Instead of writing custom validators, I now use a Claude-generated JSON repair function.

Prompt: "Write a Python function that takes malformed JSON (missing commas, extra brackets) and returns valid JSON using error recovery." It returned a function using json_repair library and try-except fallbacks.

I added it to my webhook pipeline. Now, when a partner sends broken JSON from their old CRM, my system fixes it and logs the fix. Saved me 5+ hours of manual cleanup last quarter.

5. Build CLI Tools That Non-Tech Clients Can Use

My biggest win? A command-line tool that bulk-updates WooCommerce prices based on supplier CSVs. A client was editing 200+ products by hand every Monday.

I asked Claude: "Write a Python CLI that takes two CSVs (current products and new pricing), matches by SKU, and outputs a WooCommerce batch update JSON." It included argparse, fuzzy matching, and error logging.

I packaged it with PyInstaller. Now their intern runs it in 4 minutes every week. I charge $150/month to maintain it and add features.

6. Auto-Create API Docs from Code Comments

Every time I launch a new tool, I dread writing API docs. Then I trained Claude to generate Markdown docs from Python function comments.

Prompt: "Parse this Python file, extract functions, parameters, and return values, and output clean Markdown API docs." It read my Flask routes, pulled docstrings, and formatted them with headers and code examples.

Now I update the code, run the script, and publish. Saved 8 hours on my last launch. I even monetized it — I sell a templated version on Gumroad for $49.

How much does Claude cost for these workflows?

Claude 3.5 Sonnet costs $16 per million input tokens. Output is $72 per million. In real terms:

• Generating 1,000 product descriptions: ~$1.20
• Scraping and cleaning 500 leads: ~$0.35
• Running the JSON repair function 200 times per month: ~$0.08

I stay under $20/month across all projects. Even if you pay for Playwright or hosting, the ROI is clear.

Is Claude worth it for solo operators?

Yes — but only if you focus on tasks that either save you billable hours or generate direct income. The six skills above all meet that bar.

Forget "AI helpers" that just rewrite emails. Use Claude to write code that runs unattended, creates sellable assets, or replaces $50/hour work.

The ones that failed? Image generation, full app builds, and complex debugging. Stick to scripting, data cleanup, and API glue code. That’s where Claude wins.

If you’re a solopreneur using AI to automate your business, you need real tactics — not hype. I test tools like this every week and share what works at The Operator.

Join 14,000+ builders who get tested AI strategies every Friday. No fluff. Just code, profits, and systems that run without you.

Subscribe to The Operator at theoperatorai.io