Claude Code Skills & Slash Commands (66)
A scoring study of 8,220 CLAUDE.md files from top GitHub repos finding that files started with Claude Code's /init score higher at every file size (median 8.4 vs 7.7).
A free, MIT-licensed Claude skill that asks five setup questions, then renders UI screens with loading, empty and error states, focus rings and four responsive breakpoints.
An experiment on what survives /compact: the root CLAUDE.md always came back, but invoked skill bodies were sometimes dropped.
Why one team folded its skills' reference.md files back into single SKILL.md files, and what that choice costs compared with the documented split layout.
An A/B test of nine popular agent skills (SKILL.md files) that found four of them made no measurable difference.
Why a per-package skill in a monorepo reports "Unknown skill" when the file is read through Bash, and how loading order causes it.
A Claude Code skill that analyzes your chess games and produces a post-mortem.
Day 1 of an experiment giving Claude Code a $100 budget and 30 days to turn a profit, where it chose to build Claude Code skills for freelance business tasks around one shared context file.
A Claude Code skill that has several reviewer agents adversarially critique design choices before the agent commits to them.
An open-source kit of Claude Code skills for advanced context-engineering patterns.
A Claude Code journal plugin with a /journal skill that writes glanceable session summaries to Notion.
How 107 skills, 38 agents and 15 commands added nearly 10,000 tokens of overhead before the first prompt, and how to trim it.
A Claude Code skill that recovers Kindle highlights that are blocked from normal export.
/show-me, an agent skill that makes coding agents present plans and findings as compact visuals instead of walls of text.
Whetstone, a set of 20 Claude Code skills each distilled from one real failure.
Agent skills that apply a team's coding standards in Claude Code and Codex.
Wienerdog, an open-source memory and self-improving skills system for Claude Code and Codex.
Hallmark, an anti-AI-slop design skill for Claude Code, Cursor and Codex.
A tip on how to split guidance between CLAUDE.md, skills and agents when setting up Claude Code in a real codebase.
Bookshelf, a skill that shows book quotes between Claude Code and Codex turns instead of idle terminal output.
A Claude Code skill plus MCP server for stateful, multi-turn image editing built on a Gemini image model.
A Claude Code skill plus MCP server that provisions Cloud TPUs, serves Gemma 4 with vLLM, benchmarks it and tears it down.
Agent skills that keep NestJS code consistent when built with Claude Code or Codex.
A Claude Code skill for searching royalty-free stock photos through the Pexels API.
The layer after CLAUDE.md: using skills and hooks so you point Claude at procedures instead of re-prompting, with Go repo examples.
A /ci skill that has Claude Code check a project's CI health.
Handoff, a skill that passes verified context from one Claude Code session to the next.
Better Graphs, Claude Code skills and design rules that teach agents to produce good-looking Matplotlib charts.
Lessons from building a Claude Code plugin for Manifest V3 Chrome extensions.
Three commands that give Claude Code real knowledge of your infrastructure so it stops guessing.
What belongs in commands versus skills versus agents when configuring Claude Code.
A portable setup with tools in MCP, instructions in AGENTS.md and know-how in agent skills, so you can switch tools easily.
A site for finding and installing Claude skills aimed at non-technical users.
A skill that brings Claude Code-style dynamic workflows to Codex.
A skill that reads about 20 files to learn a codebase's unwritten conventions and enforce them.
skills-for-humanity, a collection of 171 structured reasoning skills for Claude Code.
skill-insp, a skill that scores the quality of other Claude Code skills.
How to generate custom Claude Code skills from your git history and memory files.
A spec-driven development workflow for Claude Code that builds requirements, analysis and design specs before coding.
A skill with layout and icon rules that makes AI-generated AWS diagrams usable with little cleanup.
A Claude skill for spec-driven development, built as an alternative to Kiro's spec workflow.
A Claude Code skill that scores legacy Java code from 1 to 100 and modernizes it toward Java 21.
How an undocumented skillListingBudgetFraction setting limits how many skills Claude Code loads, and how to tune it.
A Claude Code and Codex skill that turns coding sessions into deliberate learning opportunities.
Full Stack HQ, a permission-first kit installing CLAUDE.md, 10 specialist agents and 28 skills with one command.
A free permission-first CLAUDE.md and agent stack that stops Claude Code from acting before asking.
MetalBear's six Agent Skills that teach Claude Code, Cursor and other SKILL.md agents when and how to use mirrord to run local code against a real Kubernetes cluster.
HappyCapy, a Claude Skills marketplace that added share analytics, leaderboards and one-click install.
A collection of academic research skills for Claude Code.
Kstack, a skill pack for monitoring and troubleshooting Kubernetes with Claude Code.
A CLAUDE.md with 13 rules that make Claude Code write safe, idiomatic Rust.
“add a small async cache to this crate”
A CLAUDE.md with 13 rules that make Claude Code write idiomatic, production-safe Go.
“add a worker that drains an order queue”
A CLAUDE.md with 13 rules for composable, SSR-ready Vue and Nuxt components.
“add a user dashboard”
A CLAUDE.md structure, refined over 35 projects, that Claude Code follows more reliably.
A CLAUDE.md with 12 rules that stop Claude Code from writing legacy Rails patterns.
A spec-first Claude workflow with agent routing and hard guardrails, explained step by step.
14 CLAUDE.md rules that fix common Rust mistakes Claude Code makes, like unwrap on user input.
Agent-evals, a Claude skill for building your own evaluations for agents.
A practical guide to writing Claude Code skills so you stop repeating the same instructions.
(Spanish) A step-by-step guide to integrating OpenSpec versioned specs with Claude Code.
An experiment writing YAML specs for agents in Claude Code, and where the quality problem moved.
A slash command that makes Claude Code write meaningful commit messages.
Building an agent skill for @ngrx/signals and running it through a full eval pipeline to prove it helps.
A code review stack combining the caveman, brooks-lint and logic-lens skills for Claude Code.
A Claude Code skill that pairs with an AI-slop detector so each /slop run calibrates the checks.
ctx, a local SQLite-backed skill that carries a workstream's context across Claude Code and Codex sessions.