askill

The portable skill layer for AI companions -- teach your AI new capabilities.

npm i -g @aman_asmuei/askill

The aman ecosystem

| Layer | Package | What it does | |-------|---------|-------------| | 1 | acore | Identity -- who the AI is | | 2 | amem | Memory -- what the AI remembers | | 3 | atool | Tools -- what the AI can do | | 4 | akit | Toolkit -- curated tool collections | | 5 | aflow | Workflows -- multi-step processes | | 6 | arule | Rules -- behavioral constraints | | 7 | askill | Skills -- deep domain expertise |

What's a skill?

A skill is a markdown file that teaches your AI deep knowledge about a specific domain. Unlike workflows (step-by-step processes), skills are capabilities -- deep expertise the AI applies contextually.

Think of it as giving your AI a senior developer's quick reference card for any topic.

Quick start

# See what's available
askill search testing

# Install a skill
askill add testing

# List your skills
askill list

# Create your own
askill create

# Import from any markdown file
askill import ~/notes/laravel-tips.md

Built-in registry

askill ships with 12 curated skills:

| Skill | What it covers | |-------|---------------| | testing | TDD, test design, property-based testing, mutation testing | | api-design | REST conventions, error handling, pagination, versioning | | security | OWASP top 10, input validation, auth patterns, secrets management | | performance | Profiling, caching strategies, lazy loading, database optimization | | code-review | What to look for, how to give actionable feedback, severity levels | | documentation | JSDoc, README patterns, architecture decision records | | git-workflow | Branching strategies, commit messages, PR best practices | | debugging | Systematic debugging, rubber duck, binary search, logging strategies | | refactoring | Code smells, extract method, single responsibility, SOLID | | database | Schema design, indexing, migrations, query optimization | | typescript | Type safety patterns, generics, utility types, strict mode | | accessibility | WCAG guidelines, semantic HTML, ARIA, keyboard navigation |

Commands

askill                    # List installed skills
askill add <name>         # Install from built-in registry
askill create [name]      # Create a custom skill interactively
askill import <path>      # Import from any markdown file
askill remove <name>      # Remove a skill
askill search <query>     # Search the registry
askill list               # List installed skills with descriptions
askill show <name>        # Show a skill's full content
askill doctor             # Health check

Skill file format

Each skill is a markdown file with YAML frontmatter:

---
name: my-skill
description: "What this skill covers"
author: your-name
version: 1.0.0
tags: [tag1, tag2]
---

# My Skill

## Approach
- Core philosophy and principles

## Patterns
- Key techniques and patterns to follow

## Anti-patterns to avoid
- Common mistakes to prevent

How it works

1. Skills are stored in ~/.askill/installed/ as individual markdown files
2. askill generates a combined ~/.askill/skills.md manifest
3. Your AI platform loads skills.md into context (via acore injection or aman-plugin)
4. The AI applies skill knowledge contextually during conversations

Creating custom skills

askill create

This walks you through:

1. Naming your skill
2. Writing a description
3. Adding tags
4. Defining the approach/philosophy
5. Listing key patterns
6. Listing anti-patterns to avoid

The result is a properly formatted skill file, installed and ready to use.

Importing skills

Have markdown notes you want to turn into a skill?

askill import ~/docs/react-patterns.md

If the file has frontmatter, it's used directly. Otherwise, you'll be asked for a name and description.

Integration with acore

Add ~/.askill/skills.md to your AI platform's system prompt injection, alongside core.md. The aman-plugin handles this automatically if you're using the full ecosystem.

License

MIT