Skills

The Skills catalog is a community-maintained collection of structured AI agent skills. Each skill defines a specific task that LLM agents excel at, complete with metadata, recommended models, example prompts, and compatibility information.

Overview

Skills bridge the gap between raw model capabilities and practical use cases. While models define what an LLM can do, skills define how to use it effectively for a specific task.


Models (what)  →  Skills (how)  →  Output (result)
claude-opus-4-6    code-review       Actionable PR feedback
gpt-5              sql-generation    Optimized SQL query
gemini-2-5-pro     security-audit    Vulnerability report

Repository Structure

Skills live in a separate repository: openmodelsrun/skills


skills/
├── skills/                    # One YAML file per skill
│   ├── code-review.yaml
│   ├── sql-generation.yaml
│   ├── security-audit.yaml
│   └── ...
├── categories/
│   └── categories.yaml        # Category definitions
├── schemas/
│   ├── skill.schema.json      # JSON Schema for validation
│   └── category.schema.json
├── validate.py                # Validation script
└── .github/workflows/
    └── validate.yml           # CI validation on PR

Skill Format

Each skill is a YAML file validated against a JSON Schema. The filename must match the id field.

Example


id: code-review
name: Code Review
description: >
  Automated code review that provides actionable feedback on code quality,
  potential bugs, performance issues, security vulnerabilities, and style
  violations.
category: development
tags:
  - code-quality
  - automation
  - pull-request
author:
  name: OpenModels Community
  github: openmodelsrun
recommended_models:
  - claude-opus-4-6
  - gpt-5
  - gemini-2-5-pro
min_context_window: 32000
modalities:
  input:
    - code
    - text
  output:
    - text
    - code
complexity: intermediate
use_cases:
  - Pull request review automation
  - Code quality gates in CI pipelines
  - Pre-commit code analysis
example_prompt: |
  Review the following code for bugs, performance, and style...
related_skills:
  - unit-test-generation
  - security-audit
compatible_tools:
  - claude-code
  - cursor
  - kiro
  - any
created_at: "2026-05-24T10:00:00.000Z"
updated_at: "2026-05-24T10:00:00.000Z"

Schema Fields

Field	Type	Required	Description
`id`	string	Yes	Unique kebab-case identifier. Must match filename.
`name`	string	Yes	Human-readable display name.
`description`	string	Yes	What the skill does and when to use it (10–2000 chars).
`category`	string	Yes	Primary category (see values below).
`tags`	string[]	Yes	1–20 searchable tags in kebab-case.
`author`	object	Yes	Author info: `name` (required), `github`, `url` (optional).
`source_url`	string	No	Link to source code or extended documentation.
`recommended_models`	string[]	Yes	Model IDs from the OpenModels registry.
`min_context_window`	integer	No	Minimum tokens needed (1000–10000000).
`modalities`	object	Yes	`input` and `output` arrays of modality values.
`complexity`	string	Yes	`beginner`, `intermediate`, or `advanced`.
`use_cases`	string[]	Yes	1–10 concrete use cases.
`example_prompt`	string	Yes	Ready-to-use prompt template (10–5000 chars).
`related_skills`	string[]	No	IDs of related skills for discovery.
`compatible_tools`	string[]	No	AI tools this works with.
`created_at`	ISO 8601	Yes	When the skill was created.
`updated_at`	ISO 8601	Yes	When the skill was last updated.

ID	Name	Description
`development`	Development	Writing, reviewing, and improving code
`writing`	Writing	Generating and transforming written content
`data`	Data	Analyzing, transforming, and visualizing data
`research`	Research	Gathering and synthesizing information
`creative`	Creative	Brainstorming and creative problem-solving
`ops`	DevOps & Infrastructure	Deployment, CI/CD, and infrastructure
`testing`	Testing & QA	Test generation and quality assurance
`security`	Security	Vulnerability detection and hardening
`productivity`	Productivity	Workflow automation and efficiency

Complexity Levels

Level	Description	Model Tier
`beginner`	Simple single-step tasks	Works with smaller/faster models
`intermediate`	Multi-step reasoning	Needs capable mid-tier models
`advanced`	Complex analysis, large context	Requires top-tier flagship models

Compatible Tools

Skills can declare compatibility with specific AI coding tools:

claude-code, cursor, github-copilot, kiro, windsurf, opencode, gemini-cli, aider, continue, cody, any

Integration with the Platform

Skills are ingested into the OpenModels platform database and served via the REST API:


GitHub (openmodelsrun/skills)
    ↓ git pull
Server filesystem
    ↓ python -m ingestion --skills-only
PostgreSQL (skills table)
    ↓ REST API
Web UI (openmodels.run/skills)

API Endpoints

GET /api/skills — List skills with filtering and pagination
GET /api/skills/:id — Get skill details

See the Skills API Reference for full documentation.

Cross-References

Skill → Models: Each skill lists recommended_models linking to the models registry
Model → Skills: The platform can show “Skills that work well with this model” on model pages
Search: Skills are included in the global search index (Cmd+K)

Validation

All skills are validated on every pull request:


cd skills
pip install -r requirements.txt
python validate.py

Checks performed:

YAML syntax
Schema conformance (required fields, types, enums, patterns)
No duplicate IDs
Filename matches the id field
Related skills reference existing IDs

Skills API Reference — REST API documentation
Adding a Skill — Step-by-step contribution guide
YAML Schemas — Registry schema reference