fabric-claude-skills

内容来源:README.md(说明文档) · 原始地址 · 查看安装指南

原始内容

Fabric Skills for Claude

Transform 250+ Fabric patterns into 7 intelligent, composable skills for Claude

License: MIT Skills Patterns


What is this?

This project bridges Daniel Miessler's Fabric prompt library with Claude's skill system. Instead of memorizing 250+ individual patterns, you get 7 intelligent skills that route to the right pattern based on what you actually ask.

Before: "Should I use analyze_paper, summarize_paper, extract_references, or extract_wisdom?"

After: "Help me understand this paper" → The skill figures it out.


Quick Start

Installation

  1. Download the skills you need from the Releases page
  2. Extract to your Claude skills directory:
    # Example for a single skill
    unzip research-intelligence.skill -d ~/.claude/skills/
    
  3. The skill is now available in Claude

Or clone the entire repository:

git clone https://github.com/spitoglou/fabric-claude-skills.git
cd fabric-claude-skills

# Copy all skills to Claude's skill directory
cp -r skills/* ~/.claude/skills/

Available Skills

Skill Patterns Size Description
research-intelligence 7 8.5 KB Papers, claims, wisdom extraction
technical-documentation 9 13 KB Architecture docs, PRDs, code explanation
writing-enhancement 6 8.7 KB Editing, proofreading, style improvement
content-transformation 7 9.1 KB Summarization, format conversion
critical-thinking 7 11 KB Argument analysis, red-teaming, root cause
learning-education 7 11 KB Tutorials, quizzes, study materials
content-creation 6 9.4 KB Newsletters, emails, presentations

Total: 49 patterns across 7 skills (~71 KB)


Skill Details

research-intelligence

Extract insights from academic content, analyze claims, and synthesize research.

Trigger phrases: "analyze this paper", "extract key insights", "summarize the research", "what are the main claims", "compile references"

Pattern Purpose
extract_wisdom Flagship insight extraction from any content
analyze_paper Deep scientific paper analysis with methodology assessment
summarize_paper Quick paper overview
analyze_claims Truth claim evaluation with fallacy detection
extract_references Citation compilation
extract_ideas Concept extraction for innovation
extract_controversial_ideas Contested viewpoints identification

technical-documentation

Create and improve technical documentation, from architecture designs to PR descriptions.

Trigger phrases: "document this", "create a design doc", "explain this code", "write a PRD", "create a diagram", "architecture overview"

Pattern Purpose
create_design_document C4 architecture documentation
refine_design_document Improve existing design docs
review_design Architecture review and scoring
create_prd Product Requirements Documents
create_user_story Agile stories with acceptance criteria
explain_code Code walkthroughs
explain_docs Transform docs into clearer versions
create_mermaid_visualization Diagram generation
write_pull_request PR descriptions

writing-enhancement

Improve writing quality, fix errors, and transform style.

Trigger phrases: "improve this writing", "fix typos", "proofread", "make this clearer", "academic style", "humanize"

Pattern Purpose
improve_writing Enhance clarity, flow, style
improve_academic_writing Scholarly tone and structure
fix_typos Spelling, grammar, punctuation
analyze_prose Writing quality analysis
humanize Technical → conversational
write_essay Essays in specific author styles

content-transformation

Convert content between formats and summarize at different levels.

Trigger phrases: "convert to markdown", "summarize in one sentence", "make this a tweet", "create flashcards", "TL;DR"

Pattern Purpose
summarize Balanced key points extraction
summarize_micro 1-2 sentence TL;DR
create_5_sentence_summary Progressive 5-level summary
convert_to_markdown Any format → Markdown
tweet Content → tweet-length
to_flashcards Content → study cards
youtube_summary Video → key points + timestamps

critical-thinking

Analyze arguments, detect biases, and evaluate claims rigorously.

Trigger phrases: "evaluate this argument", "logical fallacies", "fact check", "identify biases", "devil's advocate", "red team this"

Pattern Purpose
analyze_claims Truth claim evaluation + fallacy detection
extract_predictions Assess forecasts and predictions
analyze_mistakes Learn from errors (5 Whys)
extract_primary_problem Root cause identification
t_red_team_thinking Adversarial analysis
extract_alpha Most novel/surprising insights
create_idea_compass 4-dimensional thought organization

learning-education

Create learning materials, explain concepts, and generate study aids.

Trigger phrases: "explain this", "help me learn", "create a quiz", "tutorial for", "study guide", "teach me"

Pattern Purpose
explain_terms Glossary with analogies
explain_math Step-by-step math concepts
create_story_explanation Concept → engaging narrative
create_quiz Questions by difficulty level
create_flash_cards Spaced repetition cards
create_reading_plan Structured learning path
dialog_with_socrates Learn through questioning

content-creation

Create polished content for publication and communication.

Trigger phrases: "write a newsletter", "draft an email", "create a post", "presentation", "blog post"

Pattern Purpose
create_newsletter_entry Newsletter sections
create_formal_email Professional correspondence
create_show_intro Podcast/show openings
enrich_blog_post Blog enhancement + SEO
create_keynote TED-style presentations
create_aphorisms Curated wisdom quotes

Architecture

Each skill follows a consistent structure optimized for Claude's context window:

skill-name/
├── SKILL.md           # Routing logic + pattern selection (always loaded)
└── references/
    ├── pattern_1.md   # Full pattern (loaded on-demand)
    ├── pattern_2.md
    └── pattern_n.md

Design Principles

  1. Progressive Disclosure - SKILL.md contains routing logic; full patterns load only when needed
  2. Intent-Based Routing - Patterns grouped by user goal, not output format
  3. Decision Trees - Smart routing based on request context, not just keywords
  4. Steelman First - Critical analysis patterns present strongest arguments before critiquing
  5. Chaining Support - Skills suggest logical follow-up patterns

SKILL.md Structure

---
name: skill-name
description: What this skill does. Trigger phrases and use cases.
---

# Skill Name

## Pattern Selection
| Intent | Pattern | When to Use |
|--------|---------|-------------|
| [goal] | `pattern_name` | [context] |

## Decision Flow
[Routing logic as text or diagram]

## Pattern References
[Links to reference patterns]

## Output Guidelines
[Consistency requirements]

## Chaining Suggestions
[Logical follow-ups]

Creating New Skills

Want to add patterns or create new skills? Follow these steps:

1. Identify Pattern Clusters

Group Fabric patterns by user intent, not output format. Ask: "What goal do these patterns serve?"

2. Create the Skill Structure

mkdir -p skills/my-skill/references

3. Write SKILL.md

Keep it under 500 lines. Include:

  • YAML frontmatter with name, description, triggers
  • Pattern selection table
  • Decision flow logic
  • Output guidelines

4. Adapt Patterns

Don't copy Fabric patterns directly. Adapt them:

  • Add structured OUTPUT INSTRUCTIONS
  • Include example outputs
  • Ensure consistent formatting

5. Test Routing

Verify the skill routes correctly for various phrasings of the same intent.

Contributing

  1. Fork the repository
  2. Create a feature branch (git checkout -b skill/my-new-skill)
  3. Add your skill following the structure above
  4. Submit a pull request with:
    • Description of the skill's purpose
    • List of patterns included
    • Example use cases

Roadmap

Planned Skills

Skill Patterns Status
security-analysis ~22 🔜 Planned
meeting-productivity ~8 🔜 Planned
business-strategy ~18 💭 Considering
visualization-diagrams ~14 💭 Considering

Improvements

  • Automated testing for routing logic
  • Skill dependency management
  • Version compatibility tracking
  • Usage analytics integration

Credits & Disclaimer

This project builds on the excellent work of:

  • Fabric by Daniel Miessler - The source of 250+ battle-tested AI patterns
  • Anthropic - Claude and the skill system architecture

The patterns in this repository are adapted from Fabric's open-source collection and restructured for Claude's skill format.

Disclaimer: This is an independent project. Daniel Miessler and the Fabric team have no affiliation with, involvement in, or responsibility for this work. This project simply adapts patterns from Fabric's open-source repository into a format suitable for Claude's skill system.


License

MIT License - See LICENSE for details.

The original Fabric patterns are subject to their own licensing terms.


Links


The future of AI assistance isn't more prompts—it's smarter organization of the prompts we have.