michaelshimeles/skills

What it solves

This project provides a collection of specialized "skills" for Claude Code, allowing the AI agent to perform complex, multi-step workflows that go beyond basic coding. It solves the problem of needing consistent, high-quality outputs for specific tasks like visual regression testing, architectural enforcement, and PR cleanup, by providing the agent with on-demand instructions for these scenarios.

How it works

Each skill is a standalone folder containing a SKILL.md file. This file includes frontmatter (name and description) and detailed instructions. Claude Code loads these skills on demand when the task at hand matches the description provided in the frontmatter. Users can either copy these skills into their global Claude skills directory or project-specific directories.

Who it’s for

Developers using Claude Code who want to extend the agent's capabilities with standardized workflows for UI testing, code architecture, and pull request management.

Highlights

• Visual Proofing: Includes skills for capturing before-and-after screenshots of web pages and recording visual evidence of UI behavior during testing.
• Architectural Guidance: Enforces a two-layer service layer architecture to prevent logic duplication and "god services."
• Automated PR Refinement: Iteratively fixes PRs based on Greptile reviews until a perfect score is achieved.
• Environment Isolation: Automatically manages Git worktrees to allow multiple agents to work on the same repository in parallel without conflicts.
• Human-like Prose: A tool to remove "AI tells" and puffery from commit messages, documentation, and PR descriptions to ensure a human voice.