auto-slacker

graph LR
    A[Do manual work] --> B[Spend time automating it]
    B --> C[Save time]
    C --> D[Use saved time to automate more]
    D --> C
    C --> E[Do less work]
    E --> F[Have time to automate other things]
    F --> B

What is this?

A starter kit for systematic LLM programming. Pre-made structure with examples for how to manage and develop knowledge about LLMs and scripts that work well with them. It functions both as a repository of knowledge and as a tool/brain you use when directing LLMs.

This is a simplified version of a more complex personal system (described in distillery/structure_example). It’s meant to be forked and used, not just read.

Who is this for?

• People learning to work systematically with LLMs
• Anyone wanting structure for their LLM experiments, prompts, and insights
• Those building their own “operating system” for AI-assisted development

You can fork it and evolve it on your own, or use it during guided learning sessions.

The Core Idea

Your structure in = Your goal out

LLM work is about structure. Garbage in, garbage out. This repo demonstrates that by being well-structured itself and growing through use.

The more you use it, the more valuable it becomes. It’s a self-documenting learning tool that evolves with you.

How It Works

This repo is maintained entirely through LLM prompts. No direct file editing. Every change is an example of LLM-driven development.

This proves the concept: if you can build this tool with LLMs, you can build anything with LLMs using the same patterns.

Structure

See repo-structure.md for the complete organization and purpose of each directory.

Quick Links

• llm-lore/ - What LLMs are and aren’t (fundamental concepts)
• prompt-vault/ - Patterns that work (and anti-patterns that don’t)
• context-garden/ - Factual tidbits to remember
• distillery/ - Refined workflows and examples
• rubber-duck-brain/ - Problem-solving patterns
• script-kiddies/ - Meta-automation patterns
• token-wisdom/ - Optimization techniques

How to Use This

Fork it - Make it your own. The format is not fixed.

Populate it - As you work with LLMs, capture what you learn:

• Prompts that work → prompt-vault/good/
• Prompts that fail → prompt-vault/bad/
• Insights about LLMs → llm-lore/
• Workflows you refine → distillery/
• Facts you need to remember → context-garden/

Let it evolve - The structure can grow and change with your needs. This is a starting point, not a rigid framework.

Philosophy

• If you’re doing it more than twice, automate it
• Document by doing - every commit is an example
• Collect what works, but also what doesn’t - anti-patterns are valuable
• Iterate, don’t try to fix all things in one go
• Treat this as a living project that continuously improves
• You are the product owner, stakeholder, and director - LLMs are executors of your will

Why “auto-slacker”?

The name captures the irony: spend effort now to do less later, recursively, forever.

The subdirectories are components of an automated system for LLM work. Together they help you think less about organization and more about building.

---

This README was written by an LLM, naturally.