humanlayer/skillsPublic

AI summary: Collection of advanced Claude Code skills for building agentic loops, rewriting instructions, and diagramming architecture.

Star on GitHub

Stars: 3.6K; +261 today
Forks: 106
Watchers: 7
Open issues: 3
Open PRs: 4
Contributors: ~2
Commits: 12
Branches: 4

TypeScriptMITCreated Mar 18, 2026Last push 28d ago+514 stars this week+514 this month

Star history

since Mar 29, 2026

3.6K stars as of Sep 9, 2026, tracked back to Mar 29, 2026. Historical curve reconstructed from public GitHub event archives, calibrated to the current total.

Contribution activity

commits per day, last 52 weeks

10 commits in the last yearLessMore

Signals and awards

derived from tracked data

Rising fast
+514 stars this week
Permissive license
MIT

What skills does

The skills repository by HumanLayer provides a suite of installable capabilities designed to extend Claude Code's functionality within local codebases. It offers specialized commands that modify project configuration, such as rewriting the CLAUDE.md file with precise control blocks to enforce instruction adherence. Additionally, it includes tools to build iterated, agentic coding loops directly integrated with GitHub Actions. It can also analyze React code to narrow component prop types based on actual usage rather than mock data. The project emphasizes creating autonomous, locally runnable agent components and generating visual explanations of current topics.

This toolkit is built for developers who heavily utilize Claude Code and want to extend its capabilities with structured, agentic workflows. It requires an understanding of AI prompt engineering and local execution loops.

Instruction improvement: Rewrites CLAUDE.md using important blocks to force better adherence by the AI.
Agentic loop building: Constructs repo-local skills combined with iterated GitHub Actions workflows.
React type narrowing: Analyzes live code paths to restrict and narrow React component prop types accurately.
Control loop design: Interviews the user to design custom sensor, controller, and actuator components for the codebase.
Visual explanations: Explains complex current topics by generating concise diagrams, code-shape sketches, and HTML artifacts.
Modular installation: Allows developers to install specific skills individually via npx directly into their project.

Where teams use it

Context enforcement

Developers can run the instruction improver to ensure Claude strictly follows project-specific rules defined in CLAUDE.md.

React refactoring

Frontend engineers can narrow loose prop types by scanning actual component usage across the application.

Autonomous agents

Teams can quickly scaffold an iterated coding-agent loop that runs on a schedule via GitHub Actions.

Architecture visualization

Architects can ask Claude to explain a system, automatically generating HTML artifacts and diagrams of the code shape.

Getting started: npx skills add humanlayer/skills --skill SKILLNAME

README

main branch

skills

Claude Code skills from HumanLayer.

Installation

npx skills add humanlayer/skills --skill SKILLNAME

Available Skills

improve-claude-md

Rewrites your CLAUDE.md using <important if> blocks to improve instruction adherence.

npx skills add humanlayer/skills --skill improve-claude-md

Then in your project:

/improve-claude-md

narrow-react-prop-types

Narrows React component prop types to match live code paths instead of Storybook, test, or mock-only states.

npx skills add humanlayer/skills --skill narrow-react-prop-types

Then in your project:

/narrow-react-prop-types

build-iterated-agentic-loop

Builds a repo-local skill plus an iterated coding-agent GitHub Actions workflow, prompt, memory file, and reference templates.

npx skills add humanlayer/skills --skill build-iterated-agentic-loop

Then in your project:

/build-iterated-agentic-loop

design-control-loop

Interviews you to design an agentic control loop — sensor, controller, actuator, and disturbances — tailored to your codebase, then builds it as locally-runnable components plus a scheduled coding-agent workflow.

npx skills add humanlayer/skills --skill design-control-loop

Then in your project:

/design-control-loop

show-me

Explains the current topic with concise diagrams, code-shape sketches, and focused HTML artifacts.

npx skills add humanlayer/skills --skill show-me

Then invoke:

/show-me

View on GitHub

Recent activity

commits and pull requests

docs: sync example-skill.md copies with current narrow-react-prop-types skillOpen
#7 by ruttybob · updated 5d ago
docs: clarify agent loop skill selection and terminologyOpen
#6 by kr4chinin · updated 6d ago
add some guidanceMerged
#5 by dexhorthy · updated 28d ago
Merge pull request #5 from humanlayer/helptext
dexhorthy committed 3c26291 · 28d ago
add some guidance
dexhorthy committed 6ab9013 · 28d ago
show-me: add topology view and color-encoding guidanceOpen
#4 by adamjr36 · updated 28d ago
fix(show-me): use a platform-appropriate command to open the HTML fileOpen
#3 by JuanluTrapero · updated 29d ago
show-me: use real tree glyphs in file-tree examples
dexhorthy committed 4d8d644 · 29d ago
show-me: open HTML with Bash instead of task-artifact
dexhorthy committed d3138e5 · 1mo ago
add show-me skill
dexhorthy committed a2e3bf7 · 1mo ago
Merge branch 'main' of github.com:humanlayer/skills
dexhorthy committed 78d9dff · 1mo ago
feat: add design-control-loop skill
K-Mistele committed 39fb327 · 2mo ago

Recent open issues

view all

OrcaRouter provider support for humanlayer/skills agent loops
#8 · 4d ago · by dulcestentaciones2920-debug
design-control-loop vs build-iterated-agentic-loop?
#2 · 1mo ago · by sake92 · 1 comments
Following your install direction installs a different skill with the same name
#1 · 4mo ago · by ryankv

Commits per week

last 52 weeks

10 commits in the last 52 weeks.

When work happens

weekday and hour

Commit volume by weekday and hour (UTC). Larger dots mean more commits.

Date	List	Rank	Stars gained
Sep 7, 2026	daily	#6	+442
Sep 6, 2026	daily	#6	+442

freeCodeCamp/freeCodeCamp
freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming, and computer science for free.
455.3K stars · TypeScript
openclaw/openclaw
The AI that really does things. Any OS. Any Platform. The lobster way. 🦞
389.4K stars · TypeScript
openclaw/openclaw
Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
384.4K stars · TypeScript
openclaw/openclaw
Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
384.4K stars · TypeScript
openclaw/openclaw
Your own personal AI assistant. Any OS. Any Platform. The lobster way. 🦞
384.4K stars · TypeScript
anomalyco/opencode
The open source coding agent.
206.4K stars · TypeScript

Star history

Contribution activity

Signals and awards

What skills does

Where teams use it

Context enforcement

React refactoring

Autonomous agents

Architecture visualization

README

skills

Installation

Available Skills

improve-claude-md

narrow-react-prop-types

build-iterated-agentic-loop

design-control-loop

show-me

Recent activity

Recent open issues

Commits per week

When work happens

Trending appearances

Related repositories