Repolode
Submit
humanlayer

humanlayer/skills

A collection of Claude Code skills from HumanLayer for research, design, PR review visuals, and agentic coding workflows.

View on GitHub
Stars
5,348
Forks
177
Open issues
9
Last commit
11 days ago

What is it?

This repository is a set of installable skills for Claude Code, the coding agent from Anthropic. Each skill is a packaged instruction set that adds a slash command, such as /rpi or /show-me, and is installed with the skills CLI via npx skills add humanlayer/skills --skill NAME.

The skills cover several parts of the development loop. The rpi skill researches a change, works through design choices with the user, and implements it from local documents, with options for design discussion, PRD plus TDD, technical design, or research followed by implementation. show-me explains a topic with diagrams and HTML artifacts, and visual-pr creates or updates pull requests with a visual outline for reviewers.

Other skills target project setup and agent configuration. improve-claude-md rewrites CLAUDE.md using conditional instruction blocks, narrow-react-prop-types tightens React prop types to live code paths, and build-iterated-agentic-loop and design-control-loop generate scheduled coding-agent workflows for GitHub Actions. The rpi skill also works in Codex as $rpi.

Key features

  • 01Installs individually with npx skills add and a --skill flag.
  • 02rpi runs research, PRD, technical design, and implementation phases, each in a fresh session by default.
  • 03show-me produces concise diagrams and focused HTML artifacts.
  • 04visual-pr creates or updates pull requests with a visual change outline.
  • 05improve-claude-md restructures CLAUDE.md with conditional instruction blocks.
  • 06Workflow skills generate GitHub Actions coding-agent loops with prompts and memory files.

Use cases

  • Planning a feature with Claude Code by researching the codebase and writing a PRD before implementation.
  • Helping reviewers understand a large pull request through a visual outline.
  • Tidying an existing CLAUDE.md so the agent follows project rules more reliably.
  • Setting up a scheduled coding-agent workflow in a GitHub repository.

Quick start

  1. 1

    Install the rpi skill

    npx skills add humanlayer/skills --skill rpi
  2. 2

    Invoke the skill with your task in Claude Code

    /rpi research how the collections connect to the database

Things to know

  • Each skill is installed separately with the --skill flag.
  • Codex use needs write access to .agents/artifacts because its default sandbox can make .agents read-only.

Topics

#claude-code#ai-agents#coding-agent#prompt-engineering#github-actions#react#developer-workflow

FAQ

What is skills?
A collection of Claude Code skills from HumanLayer for research, design, PR review visuals, and agentic coding workflows. This repository is a set of installable skills for Claude Code, the coding agent from Anthropic. Each skill is a packaged instruction set that adds a slash command, such as /rpi or /show-me, and is installed with the skills CLI via npx skills add humanlayer/skills --skill NAME.
Is skills free to use? What license does it have?
skills is open source under the MIT license. Check the license terms before using it in commercial products.
Is skills still maintained?
Actively maintained. The last commit to skills was on Oct 8, 2026.
What can I use skills for?
Planning a feature with Claude Code by researching the codebase and writing a PRD before implementation. Helping reviewers understand a large pull request through a visual outline. Tidying an existing CLAUDE.md so the agent follows project rules more reliably. Setting up a scheduled coding-agent workflow in a GitHub repository.
How do I install skills?
Install the rpi skill: npx skills add humanlayer/skills --skill rpi Invoke the skill with your task in Claude Code: /rpi research how the collections connect to the database
Where can I get skills?
The source code and installation instructions are on GitHub: https://github.com/humanlayer/skills

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects