agent-skills: Reusable AI Agent Skills for Specification-Driven Development

Introduction

agent-skills is a powerful collection of reusable AI agent skills designed to streamline the entire software development lifecycle through specification-driven and autonomous development. This repository provides a comprehensive suite of tools, from generating detailed requirements and designs to automating implementation, code review, testing, and project orchestration. It aims to enhance developer productivity, ensure consistent quality, and enable advanced autonomous development workflows.

Installation

To get started with agent-skills, you can install all skills globally or choose to install individual skills as needed.

# Install all skills
npx skills add anyoneanderson/agent-skills -g -y

# Install individual skills, for example:
npx skills add anyoneanderson/agent-skills --skill spec-generator -g -y
npx skills add anyoneanderson/agent-skills --skill spec-writing -g -y
# ... and so on for other skills

Note: The cmux skills require cmux (macOS 14.0+) and must be executed within a cmux session.

Examples

Here are some quick start examples demonstrating how to use agent-skills:

Generate Specifications

> Generate requirements
> Create a design document for a todo-app
> Create a task list for a todo-app
> Generate all specifications for an e-commerce site

Resume Session Across Sessions

> handover write
> handover boot
> handover install
> handover status

Inspect Specification Quality

> Inspect specifications
> Quality check
> Validate specifications

Implement a Single Task

> /spec-code --issue 42 --task T-003 --spec .specs/auth-feature/
> /spec-code --task T-007 --feedback .specs/feature/review-T-007.md

Orchestrate the Entire Pipeline from Spec to PR

> Implement from spec --issue 42
> Start implementation --spec .specs/auth-feature/
> Resume implementation --resume

Set Up Autonomous Harness Development

Prerequisite: First, run /spec-rules-init and /spec-workflow-init. Harness consumes the coding-rules.md, review_rules.md, and issue-to-pr-workflow.md generated by these.

> Initialize harness          # harness-init: Introduces the control loop
> Plan epic               # harness-plan: Drafts product-spec ? roadmap ? tracker Issue
> Run harness-loop        # harness-loop: Autonomous Generator ? Evaluator sprint ? PR
> Run harness-loop --mode autonomous-ralph

Why Use It

agent-skills offers a comprehensive suite of tools that significantly enhance the development process by leveraging AI. It provides:

• Specification Management: Skills like spec-generator, spec-writing, and spec-inspect automate the creation, refinement, and validation of project specifications, ensuring clarity and consistency from the outset. spec-to-issue can even convert specifications into structured GitHub Issues.
• Development Automation: From spec-code for autonomous task implementation to spec-review for structured code reviews and spec-test for test creation and execution, agent-skills covers the core development cycle. spec-implement and spec-orchestrate can manage the entire pipeline from specification to Pull Request.
• Agent Collaboration: Features like handover facilitate session continuity, while cmux-fork, cmux-delegate, cmux-second-opinion, and agent-delegate enable seamless collaboration and adversarial reviews between different AI agents, even across various LLM CLIs via magi.
• Autonomous Development (Harness Engineering): For advanced users, the Harness Engineering suite (harness-init, harness-plan, harness-loop) introduces a self-driving development loop, allowing AI agents to plan epics, manage sprints, and iterate towards Pull Request creation with minimal human intervention.
• Compatibility: The skills are designed to work with a wide range of SKILL.md format-compatible agents, including Claude Code, Cursor, Codex, Gemini CLI, and OpenCode, ensuring broad applicability across different AI development environments.

Links

• GitHub Repository: https://github.com/anyoneanderson/agent-skills
• License: MIT License