I built a plugin that turns your AI chats into automation
A Claude Code Plugin that creates custom Claude Skills in 5 minutes
I’ve asked Claude to format emails the same way 47 times. Not exactly 47 — I’m guessing — but it’s a lot. Each time, I’m explaining the structure from scratch. “Make it casual but professional, lead with the question, keep it under three paragraphs.”
While I’m capable of writing a perfect prompt every time, I’m human. I’m lazy. Of those not-so-hypothetical 47 times, the quality of the prompt varied but my intent stayed the same. Asking a human to be perfect, repeatedly, is a recipe for disappointment.
The primitive that changes everything
When Anthropic announced Skills last week, I got excited. Skills are purpose-built, no-code agents with custom instructions that load dynamically when needed. They’re reusable (one skill can call another), flexible (work in Claude Code, plugins, projects, web, mobile), and omnipresent across the ecosystem.
In this way, Skills are a primitive — a building block. Primitives have to be flexible by nature, but to maximize impact they also need to be accessible to as many users as possible. The more accessible, the greater the impact. I learned this leading the Slack Platform. Anthropic’s decision to base Skills on markdown files — great for both LLMs and humans — is brilliant.
But here’s what we also learned at Slack: while people love the idea of automations they struggle knowing where to start. Systems thinking isn’t first nature to most.
Most platforms solve this with ready-to-use templates that you can also fork and make your own. Claude ships pre-defined skills in Settings that you can enable and even a “Create a skill” skill that interviews you. Clever.
But it’s still work. And I still feel like I’m repeating myself.
The best candidates already exist
I’m sure I’ll write Skills from scratch. But it’s hard the first time you do something to know how many times you’ll use it again. Is this repeatable? Should I make this a workflow? How many times will I use it?
The irony: we’re in the best position to answer these questions *after* using something several times, not at creation time.
The best candidates for my Skills already exist — in my conversation history. I just needed a way to identify the patterns, validate their significance, and codify them into reusable Skills. Actually, I wanted Claude to extract them.
So I built a Claude Code plugin that does exactly that.
What it actually does
The analyzer reads your exported conversations from both Claude and ChatGPT, performs pattern analysis across your entire history, and automatically generates Custom Skills packages.
It doesn’t just count how many times you’ve asked for something. It performs:
Temporal analysis to understand usage trends
Complexity assessment to evaluate automation potential
Cross-platform deduplication to recognize when you started work in ChatGPT and continued it in Claude or vice versa
Quality scoring to prioritize high-impact patterns
The output: Complete, ready-to-use Skills. Each includes the main skill file with proper metadata (so Claude knows when to use it), detailed methodology, usage examples from your actual conversations, and reusable templates.
Voilà. Codified versions of your actual workflows, extracted from how you already work.
The recursive loop
The plugin analyzes AI conversations to improve AI conversations. Your usage patterns train the analyzer. The analyzer generates skills. Claude uses those skills to improve output quality. Better outputs refine your patterns. The cycle continues.
This isn’t just a cool idea — the plugin makes it practical through incremental processing. After the first analysis, it maintains a log of what it’s already processed. When you export new conversations next month, it only analyzes the new material. Your history keeps growing, but you’re not reprocessing everything every time.
Each iteration codifies more of your institutional knowledge into reusable, shareable automation.
How to get started
If you are already familiar with Claude Code — run these commands from a new project folder:
# Add the marketplace
/plugin marketplace add hirefrank/hirefrank-marketplace
# Install the analyzer
/plugin install claude-skills-analyzer@hirefrank
# Complete setup walkthrough
/skills-setupIf you are new to Claude Code read the detailed instructions.
🙏If you try it, I’d genuinely love to hear what you find in the comments below. What did it get right? What did it miss? What broke?
Privacy note: All processing happens locally on your machine. Your conversation exports never leave your computer. The generated skills automatically remove sensitive information.
—
GitHub: hirefrank/claude-plugins
Anthropic: What are Skills?
Anthropic: How to create custom Skills