agent-skills-sample/README.md
Hofmann 1958f8e17e Initial commit: 7-Skill System for Claude Code
Starter template with requirements, architecture, frontend, backend,
qa, deploy, and help skills — inspired by Alex Sprogis' guide.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-01 19:12:09 +01:00

2.3 KiB

Agent Skills Sample

A starter template demonstrating the 7-Skill System for Claude Code, inspired by Alex Sprogis' guide.

What Are Agent Skills?

Agent Skills are markdown files that give AI coding agents procedural knowledge. One file, and your agent works like a specialist instead of a generalist.

.claude/skills/
├── requirements/SKILL.md   # Product Manager
├── architecture/SKILL.md   # Tech Lead
├── frontend/SKILL.md       # Frontend Dev (sub-agent)
├── backend/SKILL.md        # Backend Dev (sub-agent)
├── qa/SKILL.md             # QA Engineer (sub-agent)
├── deploy/SKILL.md         # DevOps
└── help/SKILL.md           # Guide

Quick Start

# Clone
git clone https://git.hofmanns.tech/hofmann/agent-skills-sample.git
cd agent-skills-sample

# Install
npm install

# Start building with skills
claude
> /help                    # Learn the workflow
> /requirements login      # Define a feature
> /architecture PROJ-001   # Design it
> /frontend PROJ-001       # Build the UI
> /backend PROJ-001        # Build the API
> /qa PROJ-001             # Test it
> /deploy production       # Ship it

Creating Your Own Skills

---
name: my-skill
description: What it does and when to trigger it
disable-model-invocation: true   # User-only (for side effects)
context: fork                     # Run in isolated sub-agent
allowed-tools: Read, Write, Bash  # Restrict tool access
---

# Skill Instructions

Your instructions here. Use $ARGUMENTS for user input.
Use !`command` for dynamic context injection.

Key Concepts

Concept Description
$ARGUMENTS User input after skill name: /deploy staging -> "staging"
context: fork Runs in isolated sub-agent (own context window)
disable-model-invocation Only user can trigger (prevents accidental runs)
!command`` Injects command output before skill runs
allowed-tools Restricts which tools the skill can use

Resources

License

MIT