Skip to content

Latest commit

 

History

13 Commits

Folders and files

NameName
Last commit message
Last commit date
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Skill Flash

Tools for optimizing and integrating AI skills across multiple platforms.

Overview

Skill Flash provides two complementary skills for managing AI agent capabilities:

  • skill-compilation - Optimize skills into efficient three-tier architecture
  • skill-burning - Intelligently integrate skills into project behavior

Philosophy

慢编译、快执行

Complex analysis upfront → Fast, accurate execution thereafter

Just like compiled languages (C++, Go) trade slower compilation for faster execution, these skills invest time in analysis to deliver optimal runtime performance.

Skills

skill-compilation

Optimize AI skills by restructuring into efficient three-tier architecture.

Benefits:

  • Reduce token usage by 70%+
  • Faster skill execution
  • Automatic script generation
  • Cross-platform conversion

Use when:

  • A skill is too long / uses too many tokens
  • You want faster skill execution
  • You want to convert skills between platforms

Output:

  • Tier 1: Core principles for system prompt (zero cost)
  • Tier 2: Session configuration (one-time load)
  • Tier 3: Optimized skill file (on-demand)
  • Scripts: Executable rules

skill-burning

Intelligently integrate skill principles into project behavior patterns.

Benefits:

  • Skills become automatic (zero invocation cost)
  • Seamless integration with existing guidelines
  • Consistent behavior across platforms
  • Permanent until rollback

Use when:

  • You want skills to be "always on"
  • You want team standards to be automatic
  • You want to avoid repeated skill loading

Output:

  • Updated project config files
  • Principles blended with existing guidelines
  • Natural integration (not simple append)

Supported Platforms

Platform Config Files Status
Claude CLAUDE.md, SKILL.md ✅ Full support
OpenAI GPTs configuration ✅ Full support
Cursor .cursorrules ✅ Full support
Copilot .copilot-instructions.md ✅ Full support

Quick Start

Compile a Skill

You: "Optimize the TDD skill"

The skill will:
1. Analyze the skill structure
2. Extract core principles (Tier 1)
3. Generate session config (Tier 2)
4. Create optimized skill (Tier 3)
5. Generate executable scripts
6. Output with quality report

Burn a Skill

You: "Burn TDD principles into this project"

The skill will:
1. Analyze existing project guidelines
2. Extract TDD core principles
3. Find natural integration points
4. Match project style and format
5. Update config files seamlessly
6. Verify quality and consistency

Workflow

Original Skill
      ↓
skill-compilation (optimize structure)
      ↓
Core Principles + Optimized Skill
      ↓
skill-burning (integrate into project)
      ↓
Automatic Behavior

Combine both:

  1. Compile skill → extract core principles
  2. Burn principles → integrate into project
  3. Result → Skills are automatic, zero cost

Directory Structure

skill-flash/
├── README.md                    # This file
├── .gitignore
│
├── skill-compilation/           # Compilation tool
│   ├── README.md
│   ├── SKILL.md                 # Main guide
│   ├── examples/                # Complete examples
│   ├── specs/                   # Technical specs
│   │   ├── platforms/           # Platform guides
│   │   ├── code-examples.py     # Script templates
│   │   └── quality-checklist.md
│   ├── references/              # Detailed docs
│   └── templates/               # Output templates
│
└── skill-burning/               # Burning tool
    ├── README.md
    ├── SKILL.md                 # Main guide
    ├── examples/                # Complete examples
    ├── specs/                   # Technical specs
    │   ├── platforms/           # Platform guides
    │   ├── burning-strategies.md
    │   └── verification.md
    └── templates/               # Integration examples

Examples

See examples/ directories for:

  • skill-compilation/examples/tdd-compilation.md
  • skill-burning/examples/tdd-burning.md

Quality Assurance

Both skills include comprehensive quality mechanisms:

  • Coverage checks - Ensure no content is lost
  • Style matching - Blend naturally with existing content
  • Verification - Consistency and flow checks
  • Rollback - Safe operations with backups

Contributing

These skills are designed to be executed by LLMs. To improve:

  1. Test with real projects and skills
  2. Note where the LLM gets confused
  3. Clarify those sections
  4. Add more examples
  5. Expand platform support

License

MIT

Repository

https://github.com/Army1900/skill-flash

About

compilation and burning skills

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages