Structured framework knowledge for AI coding assistants: best practices, patterns, and workflows.
Note
AI Authorship Disclosure: Drafted with GPT-5.3 Codex from official documentation, then reviewed, refined, and validated by humans. Tested with AI assistants and improved through practical iteration.
AI coding assistants are powerful, but they often:
- Generate outdated framework patterns and anti-patterns
- Miss critical security checks (permissions, validation, authentication)
- Skip proper architectural patterns and conventions
- Ignore existing tooling and best practices in your project
Agent Skills solve this by giving AI assistants deep framework knowledge in a structured, discoverable format.
|
|
The easiest way to install is using the Skills CLI:
# Install all skills from all frameworks
npx skills add lubusIN/agent-skills
# Install all skills from all frameworks globally
npx skills add lubusIN/agent-skills -g
# Install all Frappe skills
npx skills add lubusIN/agent-skills --skills=frappe-*
# Install specific skills from any framework
npx skills add lubusIN/agent-skills --skills=frappe-doctype-development,frappe-api-developmentThe CLI automatically detects your AI assistant (Claude Code, Cursor, Codex, VS Code) and installs to the correct location.
Refer Skills CLI doc for all the available args
If you prefer manual installation:
# Clone this repo
git clone https://github.com/lubusIN/agent-skills.git
cd agent-skills
# Copy all Frappe skills to your AI assistant's skills directory
# Claude Code (global)
cp -r skills/frappe/frappe-* ~/.claude/skills/
# Cursor (global)
cp -r skills/frappe/frappe-* ~/.cursor/skills/
# Or into your project
cp -r skills/frappe/frappe-* /path/to/your-project/.claude/skills/Each skill contains:
skills/frappe/frappe-doctype-development/
├── SKILL.md # Main instructions (when to use, procedure, verification)
└── references/ # Deep-dive docs on specific topics
├── doctypes.md
├── controllers.md
└── ...
When you ask your AI assistant to work on framework code, it reads these skills and follows documented procedures rather than guessing.
Each SKILL.md follows a standard format:
- When to use - Trigger conditions for this skill
- Inputs required - What info the agent needs before starting
- Procedure - Step-by-step instructions
- Verification - Checklist to confirm success
- Failure modes - Common issues and fixes
- Escalation - When to consult docs or ask the user
Contributions welcome! To add or improve a skill:
- Fork this repository
- Create or modify skill in
skills/<topic>/<skill-name>/ - Ensure SKILL.md has all required sections
- Test with your AI assistant
- Submit a pull request
- Keep
SKILL.mdshort and procedural - Push detailed content into
references/ - Include verification steps
- Document failure modes
- Link to official docs when appropriate
- Create
skills/<topic>/directory - Add
skills/<topic>/README.mdwith topic-specific documentation - Create individual skill directories following the Agent Skills specification
- Update the main README.md to list the new topic
- Inspired by WordPress Agent Skills
- Built on the Anthropic Agent Skills Specification
LUBUS is a web design agency based in Mumbai.
Agent Skills is open-sourced licensed under the MIT License.