Turn any knowledge source into an automated Claude Code workflow — like loading "kung fu" into your AI assistant's brain.
This is a method for turning Claude Code into a virtual employee that performs complex, repeatable workflows automatically. It combines two mechanisms:
Expert knowledge loaded from any source — transcripts, tutorials, documentation, best practices — saved as markdown files that Claude Code can read on demand.
A persistent instruction file that tells Claude when and how to use each skill — the routing layer that ties everything together into an automated workflow.
Just like Neo loading kung fu into his brain — you're loading expert-level knowledge into Claude Code. It then knows how to execute that skill better than you can manually prompt it, because the entire methodology is embedded in context every time.
After a /clear in Claude Code, the conversation is completely fresh — except for what's in claude.md and your loaded skills. This means every new session starts with a clean context window but retains all your pre-built workflows. Claude Code also happens to be exceptionally good at writing prompts, so when it uses a skill (like prompting an image generator), it often performs better than you would manually.
Take expert knowledge from any source and save it as a .md file. The creator's example: grabbing YouTube video transcripts about thumbnail design rules and saving them as a skill.
Sources for skills: YouTube transcripts, documentation, tutorial notes, style guides, SOPs, playbooks, research papers — anything with structured expert knowledge.
This is the routing layer that tells Claude Code when to activate each skill and how to chain them together.
With skills loaded and memory configured, every new Claude Code session automatically has access to your workflows — even after /clear.
The system is designed for rapid iteration. Example follow-ups:
Fill in the fields below and we'll generate both your skill.md file and your claude.md memory routing — ready to paste into Claude Code.
Use this as your first message after /clear to activate the workflow:
The creator's example was YouTube thumbnails, but this pattern works for any repeatable workflow. Here are concrete applications:
Load WordPress development best practices as a skill. Memory triggers on "build me a site about X" and chains theme selection → page structure → plugin configuration → content generation.
Load UI/UX design principles + your brand guidelines. Trigger with "design a page for X" and Claude generates image mockups → HTML/CSS/React code from the designs.
Load SEO writing methodology + brand voice guide. "Write an article about X" triggers keyword research → outline → draft → optimization → meta tags → internal linking suggestions.
Load technical SEO checklist + local SEO methodology. "Audit this site" triggers crawl analysis → on-page review → schema recommendations → competitor gap analysis → action plan.
Load copywriting frameworks + objection handling. "Create outreach for X prospect" generates personalized email sequences with A/B variants and follow-up cadences.
Chain multiple skills: script writing → shot list → thumbnail → title → description → tags. The creator actually got Claude Code to edit an entire video using this approach.
Every use case follows the same formula: