I run 3 content sites, 51 APIs, and 10 Chrome extensions from Paraguay — alone.
Not because I'm exceptional at time management. Because I've spent 6 months finding the Claude prompts that replace hours of work with minutes of work.
Here are the 5 categories that generate the most leverage, with real examples.
Category 1: Audit Prompts (Replace Reviews)
The pattern: ask Claude to review something as a specific expert who would catch specific problems.
The prompt structure:
You are a [specific expert] reviewing [artifact] for [specific failure mode].
Your job is to find problems, not validate the work.
Review this [artifact] and list every issue you find, ordered by severity.
[paste artifact]
Real example (SEO audit):
You are an SEO specialist reviewing a blog post for issues that would prevent it from ranking.
Your job is to find problems, not validate the work.
Review this post and list every SEO issue ordered by severity: [post]
This replaced 30-minute manual audits with 3-minute reviews that catch things I'd miss.
Category 2: Transformer Prompts (Change Format Without Losing Information)
The pattern: convert existing content to a new format while preserving the key points.
Real examples:
- Blog post → Twitter thread (with hook + 7 tweet breakdown)
- Technical doc → 5-bullet executive summary
- Meeting notes → action items with owners and deadlines
- Code comment → inline documentation
- Changelog → user-facing release notes
Key prompt element: always specify "preserve all technical accuracy" or you'll get paraphrased summaries that lose important details.
Category 3: Generator Prompts (Create From Constraints)
The pattern: provide tight constraints and let Claude generate within them.
Without constraints, you get generic output. With constraints, you get something usable on the first pass.
Constraint types that work:
- Specific word count
- Named persona ("write as someone who has actually done this")
- Required elements ("must include one specific data point and one personal anecdote")
- Format template ("use this exact structure: [template]")
- Prohibition list ("don't use the word 'leverage', don't start with 'In today's world'")
Category 4: Debugger Prompts (Find Root Causes)
Developers know you can use Claude to debug code. But the highest-leverage debugging prompts aren't "fix this error" — they're "explain why this happens."
The pattern:
Here is some code that does [expected behavior].
Here is the actual behavior I'm seeing: [actual behavior]
Don't fix it yet. First explain: why does this happen? What's the root cause?
Getting the root cause explanation first produces better fixes than jumping straight to the solution.
Category 5: System Prompts (Build Reusable Workflows)
The highest-leverage prompts aren't one-off — they're systems you reuse.
A system prompt defines Claude's role, constraints, and output format. Combined with a data input, it produces consistent output without re-prompting from scratch.
Real system prompt I use for article generation:
You are a blog writer for a site about [topic].
Rules:
- Write from first person as someone who has personally done this
- Use specific numbers when possible ("3 weeks" not "a few weeks")
- No filler phrases: "It's worth noting", "In conclusion", "As you can see"
- Short sentences. Max 20 words per sentence.
- Every section needs at least one specific example
Input: [keyword + angle + any personal experience notes]
Output: Full article draft in markdown
Reusable system prompts are worth more than individual clever prompts.
The Complete Prompt Library
I've compiled 50 of these into the Claude Code Automation Playbook — organized by use case (content, code, business, research) with real examples for each.
It's $7. I built it because I kept losing my best prompts in chat history.
Writing from Asunción, Paraguay. Using AI to run a portfolio of projects from a laptop. Follow for what's actually working.
Top comments (0)