Bootstrap 31 Cross-Agent Coding Skills with OK Skills

Planning files, docs lookup, browser automation — you can re-prompt an agent for all three in every project, or install them once as skills. ok-skills, a 455-star curated collection mentioned in Awesome Codex CLI, bundles 31 skill directories you clone in one command, plus a playbook for wiring them into your agent config.

Why This Skill Matters

The repo collects curated AI coding agent skills and CLAUDE.md / AGENTS.md playbooks for Codex, Claude Code, Cursor, OpenClaw, Autohand Code, Trae, and other SKILL.md-compatible tools. The "31 reusable skills" claim in the README holds up: the repository carries exactly 31 top-level skill directories (counted September 15, 2026), all maintained in-repo.

The README's start-here trio covers the highest-leverage habits: planning-with-files (persistent planning with task_plan.md, findings.md, and progress.md for anything that will take 5+ tool calls), find-docs (current library docs and API references via the Context7 CLI), and agent-browser (browser automation for screenshots, forms, scraping, and web QA). The wider index spans research and docs, planning and prompting, automation and QA, and frontend design; the README's popular-use-case line also lists codebase understanding and authoring for PDF, Word, PPTX, and XLSX.

For global instructions, CLAUDE_AGENTS.md packages a Chinese-first, KISS-oriented workflow — current-doc and source checks, a multi-agent preference, planning-with-files and karpathy-guidelines defaults, strict TypeScript expectations, and context-mode routing rules — meant to be copied or merged into Claude Code's CLAUDE.md or Codex's AGENTS.md.

Installation

The documented quick start clones the repo into ~/.agents/skills:

mkdir -p ~/.agents/skills
cd ~/.agents/skills
git clone https://github.com/mxyhi/ok-skills.git ok-skills

For Autohand Code, the README documents user-level (~/.autohand/skills) and project-level (.autohand/skills) clones instead.

Real Workflow: Wire the Triggers into Your Agent

1. After cloning, add the README's trigger rules to your AGENTS.md — or merge the same rules into CLAUDE.md:

## Skills

- planning-with-files: Use for complex tasks, research, or anything that will take 5+ tool calls.
- find-docs: Use when you need current library docs, API references, or Context7-backed examples.
- agent-browser: Use for browser automation, screenshots, scraping, web testing, or form filling.

2. Then ask naturally, the way the README puts it: Use planning-with-files before refactoring this module., Use find-docs to fetch the latest docs for this SDK., or Use agent-browser to reproduce this UI bug.
3. For Claude Code or Codex global instructions, start from CLAUDE_AGENTS.md, copy or merge it into CLAUDE.md / AGENTS.md, and edit its 语言要求 section for your project before reuse.

Tips

• Many skills are vendored with their upstreams recorded — planning-with-files tracks OthmanAdi/planning-with-files as its canonical baseline, and the index's Source URL column points at each upstream when a skill is imported.
• Check prerequisites per skill: some assume gh installed and authenticated; browser-focused skills may require a Chrome/CDP-capable environment; diagram-design PNG export needs Playwright and Chromium (pip install playwright && playwright install chromium).
• Contributing rules keep triggers honest: explicit and testable trigger conditions, concise execution-oriented examples, and external dependencies documented in SKILL.md.

When Not to Use This

The playbook is Chinese-first — fine for Chinese projects, but edit the 语言要求 section before adopting it elsewhere. The collection deliberately mirrors many upstreams into one place; if you depend on a single skill's evolution, follow its Source URL and track the upstream directly. And the clone-based install puts 31 directories on your machine — if you only need one or two, copy those directories instead of wiring triggers for all of them.

See the leaderboard for more skills.