🧠 OpenSkills

49 production-grade AI engineering skills for Markdown-aware agents.

github.com/CODE-SAURABH/OpenSkills

openskills is an open-source collection of 49 production-grade AI agent skills. The skills are plain Markdown and can be used with any agent that lets you load instructions.

Each skill is a single SKILL.md file that gives your AI agent opinionated, expert-level instructions for a specific engineering role — from system design and backend dev to security, QA, and retrospectives. Your agent must be configured to load the file or rule; copying a file alone does not guarantee that every product will activate it.

Start with one skill, confirm that your agent loads it, then add the rest you need.

Table of Contents

• Why OpenSkills?
• Quick Install
• The Sprint Flow
• Skills
• Agent Compatibility
• How Skills Work
• Contributing
• License

Why OpenSkills?

Most AI coding assistants are generalists. They write code, but they don't ship like a senior engineer. They don't ask the hard questions before building, they don't catch security holes, they don't enforce architecture decisions, and they don't know when to stop.

OpenSkills fixes this. Each skill turns your agent into a domain expert with:

• ✅ Explicit trigger conditions — the agent knows when to activate
• ✅ Numbered execution workflows — step-by-step, not a pile of tips
• ✅ Output templates — structured, consistent deliverables every time
• ✅ Definition of Done checklists — verifiable completion criteria
• ✅ Edge cases and anti-patterns — the things juniors miss

Quick Install

Claude Code

Clone this repository, then copy the individual skill folder you want into Claude Code's skills directory. Do not copy the entire repository: each installed skill should contain its own SKILL.md file.

git clone https://github.com/CODE-SAURABH/OpenSkills.git openskills
cd openskills

For example, install the code-reviewer skill:

mkdir -p ~/.claude/skills
cp -r code-reviewer ~/.claude/skills/code-reviewer

To install another skill, substitute its directory name:

cp -r <skill-name> ~/.claude/skills/<skill-name>

Start a new Claude Code session after copying the skill. It will then be available for use.

Codex

Codex discovers skills in .agents/skills/ within a repository and in ~/.agents/skills/ for your personal, cross-project skills. Copy a selected OpenSkills folder to one of those locations.

# Install for every project you use with Codex
mkdir -p ~/.agents/skills
cp -r code-reviewer ~/.agents/skills/code-reviewer

# Or, from a target project's root, install for that project only
mkdir -p .agents/skills
cp -r /path/to/openskills/code-reviewer .agents/skills/code-reviewer

Codex detects skill changes automatically. Mention the skill explicitly with $code-reviewer when you want to ensure it is used, or let Codex choose it when the task matches the skill description. See the Codex skills documentation for details.

Antigravity

Antigravity uses the same workspace skill layout: <workspace-root>/.agents/skills/<skill-name>/SKILL.md. From the root of the workspace you opened in Antigravity, copy the skill you want:

mkdir -p .agents/skills
cp -r /path/to/openskills/code-reviewer .agents/skills/code-reviewer

Start a new conversation after adding the skill. For personal skills shared across Antigravity workspaces, use ~/.gemini/config/skills/<skill-name>/ instead. See the Antigravity skills documentation.

Custom agents

Use the same folder structure when your agent supports the Agent Skills format:

<agent-config-root>/skills/
└── code-reviewer/
    └── SKILL.md

Configure the agent to list the available SKILL.md files at session start and load the selected file when its description matches the task. If your agent does not support skill discovery, read the chosen SKILL.md and pass its contents as the agent's system instructions or task context.

Other agents

For Cursor, Windsurf, GitHub Copilot, Kiro, Codex, and other agents, use that product's documented instruction mechanism and add only the skills that fit your workflow. A safe universal approach is to provide the contents of <skill-name>/SKILL.md as agent instructions or context. Do not concatenate all 49 skills into one instruction file: they can conflict and consume unnecessary context.

The Sprint Flow

OpenSkills maps to a complete engineering sprint. Use the right skill at each phase and hand off deliberately.

Skills

All 49 skills, ready to use.

Agent Compatibility

How Skills Work

Every skill is a single SKILL.md file inside a named directory. The file has two parts: a YAML frontmatter block that tells the agent when to load the skill, and a body that tells the agent exactly how to execute it.

openskills/
├── office-hours/
│   └── SKILL.md
├── backend-dev/
│   └── SKILL.md
├── code-reviewer/
│   └── SKILL.md
└── ...

SKILL.md Format

---
name: your-skill-name
description: >
  One or two sentences describing what this skill does AND when to use it.
  Make trigger conditions explicit: "Use when the user asks to X, Y, or Z."
---

# Skill Title

## When to Invoke This Skill
[Explicit trigger conditions — phrasing the user might say]

## Execution Workflow

### Step 1: Do This First
...

### Step 2: Then Do This
...

## Output Template
[Structured output the agent produces]

## Definition of Done
- [ ] Criterion 1
- [ ] Criterion 2

What Makes a Good Skill

Contributing

Contributions are welcome. Before opening a PR, please read CONTRIBUTING.md in full.

Adding a New Skill

1. Copy SKILL-TEMPLATE.md into a new directory: your-skill-name/SKILL.md
2. Fill in every section — no placeholder text
3. Self-review against SKILL-AUTHORING-CHECKLIST.md
4. Add your skill to skills.json with the correct category, phase, and tags
5. Add a row to the Skills table in this README
6. Open a PR — the checklist will be reviewed

Quality Bar

Every skill must pass these gates before merge:

• YAML frontmatter name and description are present and accurate
• description contains explicit trigger phrases the agent can match
• Numbered ## Execution Workflow section is present
• ## Definition of Done checklist is present and verifiable
• No placeholder text or TODO comments remain
• Skill is 200–600 lines (longer = not focused enough)

Project Structure

openskills/
├── README.md                           ← You are here
├── CONTRIBUTING.md                     ← Contribution guidelines
├── SKILL-TEMPLATE.md                   ← Template for new skills
├── SKILL-AUTHORING-CHECKLIST.md        ← Quality checklist for skill authors
├── skills.json                         ← Machine-readable skill manifest
├── install.sh                          ← One-command installer
├── agent-memory/SKILL.md               ← Session memory, browser QA, multi-agent
├── ai-engineer/SKILL.md                ← LLM integration, RAG, embeddings, evals
├── api-design/SKILL.md                 ← REST/GraphQL API design with OpenAPI
├── autoplan/SKILL.md                   ← Automated full planning review pipeline
├── backend-dev/SKILL.md                ← Production-grade server-side development
├── benchmark/SKILL.md                  ← Performance baseline and comparison
├── brand-guidelines/SKILL.md           ← Brand voice, tone, and visual consistency
├── code-reviewer/SKILL.md              ← Pre-merge 6-pass code review
├── commit-message/SKILL.md             ← Conventional commit messages
├── data-engineer/SKILL.md              ← ETL, streaming, data quality
├── database-architect/SKILL.md         ← Schema design, indexing, migrations
├── debugger/SKILL.md                   ← Systematic bug isolation and fix
├── design-explorer/SKILL.md            ← Generative design direction exploration
├── devex-engineer/SKILL.md             ← Developer experience improvements
├── devops-platform/SKILL.md            ← CI/CD, IaC, containers, cloud
├── doc-coauthoring/SKILL.md            ← Co-author PRDs, RFCs, and design docs
├── forward-deployment-engineer/SKILL.md ← Enterprise field deployment
├── frontend-design/SKILL.md            ← Distinctive, intentional UI design
├── internal-comms/SKILL.md             ← Incident reports, announcements, updates
├── land-and-deploy/SKILL.md            ← Merge, deploy, and verify production
├── mcp-builder/SKILL.md                ← Build MCP servers for AI agents
├── office-hours/SKILL.md               ← Pre-code idea validation
├── pdf/SKILL.md                        ← PDF read, extract, create, manipulate
├── performance-optimizer/SKILL.md      ← Profiling, caching, query optimisation
├── plan-ceo-review/SKILL.md            ← Strategic scope and value review
├── plan-design-review/SKILL.md         ← UI/UX plan rating and improvement
├── plan-eng-review/SKILL.md            ← Engineering plan stress-test
├── pptx/SKILL.md                       ← PowerPoint deck creation and editing
├── product-engineer/SKILL.md           ← End-to-end feature implementation
├── prompt-engineer/SKILL.md            ← LLM prompt design and evaluation
├── qa-engineer/SKILL.md                ← Live QA testing with browser
├── release-engineer/SKILL.md           ← Release packaging and deployment
├── research-agent/SKILL.md             ← Deep technical research
├── retro-engineer/SKILL.md             ← Sprint retrospectives
├── safety-guardrails/SKILL.md          ← Guardrails for destructive commands
├── security-engineer/SKILL.md          ← AppSec, OWASP, threat modelling
├── skill-creator/SKILL.md              ← Meta-skill: build and improve skills
├── spec-author/SKILL.md                ← Requirements writing
├── sre-canary/SKILL.md                 ← Canary deployment monitoring
├── system-design/SKILL.md              ← Systems architecture and MBSE
├── technical-writer/SKILL.md           ← Docs, runbooks, ADRs
├── test-writer/SKILL.md                ← Unit, integration, and E2E tests
├── theme-factory/SKILL.md              ← Design themes and style systems
├── webapp-testing/SKILL.md             ← Playwright E2E browser testing
└── xlsx/SKILL.md                       ← Excel spreadsheet engineering

License

MIT © CODE-SAURABH

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

Built for engineers who want their AI to meet their bar — not lower it.

🤝 Contribute a skill