name

skill-extraction

description

Use when user wants to convert a document/file into a reusable skill, has a methodology or framework in a file that should become actionable, asks to "make this into a skill" or "extract a skill from this", or when analyzing methodology documents, framework guides, or procedural content for skill creation. Invoke when user mentions creating skills from files, extracting methodologies, or converting documentation into structured workflows.

Skill Extraction Framework

Purpose
When to Use This Skill
What is Skill Extraction?
Workflow
Quick Reference

Purpose

Extract structured, reusable skills from documents, methodologies, frameworks, or code files. Transform written knowledge into actionable Claude Code skills that others can invoke consistently.

When to Use This Skill

User provides a file containing a methodology or framework to extract
User asks to "make this into a skill" or "create a skill from this document"
Analyzing technical documentation, guides, or procedures for skill-worthiness
Converting existing knowledge bases into executable workflows
Extracting patterns from code, notebooks, or configuration files
Building skills from research papers, blog posts, or educational content

Trigger phrases: "make this a skill", "extract skill from", "turn this into", "create skill", "methodology to skill"

What is Skill Extraction?

A systematic process for reading files using Adler's analytical reading framework adapted for LLMs to:

Understand content structure and purpose without context dilution
Identify if it contains skill-worthy methodology
Extract key components systematically
Synthesize into properly structured Claude Code skills

Quick Example:

Input: PDF about "Design Thinking methodology"
Output: design-thinking skill with 5-stage workflow (empathize, define, ideate, prototype, test)

Workflow

Copy this checklist and track your progress:

Skill Extraction Progress:
- [ ] Step 1: Pre-flight and skim
- [ ] Step 2: Classify and assess
- [ ] Step 3: Deep read and extract
- [ ] Step 4: Synthesize components
- [ ] Step 5: Generate skill files
- [ ] Step 6: Validate quality

Step 1: Pre-flight and skim

Gather file metadata (name, extension, size) and perform inspectional reading using boundary reading, structural extraction, and sampling techniques. Build mental model without reading everything. See resources/reading-strategies.md for skim techniques and resources/file-type-strategies.md for file-specific approaches.

Step 2: Classify and assess

Determine content type (methodology/framework/tool/reference), identify main theme, and assess if skill-worthy using classification criteria. Apply "What of it?" question to identify practical capability. If multiple related files exist, use multi-file analysis strategy from resources/reading-strategies.md.

Step 3: Deep read and extract

Read file systematically using section-based chunking, windowing, or targeted reading. Extract components using structured notes template. For large files (>500 lines), apply context management strategies. See resources/reading-strategies.md for deep reading techniques, resources/component-extraction.md for extraction framework, and resources/context-management.md for managing large files.

Step 4: Synthesize components

Review extracted components from notes and map to skill structure using component-to-skill mapping table. Abstract domain-specific language to generalizable patterns. Determine skill complexity and required resources. See resources/component-extraction.md for mapping guidance and pattern abstraction techniques.

Step 5: Generate skill files

Create SKILL.md with workflow, topic-specific resource files as needed, and rubric JSON following skill structure guidelines. Use progressive disclosure pattern. Reference resources/skill-structure-guidelines.md for complete structure requirements, file size limits, naming conventions, and YAML description format.

Step 6: Validate quality

Self-assess using resources/evaluators/rubric_skill_extraction.json. Check skill-worthiness, distinctness, completeness, clarity, and actionability. Minimum standard: Average score ≥ 3.5 AND no criterion < 3.0. Revise lowest-scoring criteria before delivering.