pydantic-ai-skills

What it is and what it does

pydantic-ai-skills extends agents with a structured skill system that organizes instructions, resources, and scripts into reusable packages. Each skill is a folder containing a SKILL.md file with metadata and instructions, plus optional reference documents and executable scripts. The package implements progressive disclosure: agents initially see only a skill's name and description, loading full instructions and resources on demand, which keeps the prompt lean as your skill library grows.

You can organize skills on the filesystem, define them programmatically with decorators, or load them from remote registries like Git repositories or S3. The package exposes four tools to agents: list_skills(), load_skill(), read_skill_resource(), and run_skill_script(). It supports skill filtering, runtime reloading, and composition operations like combining or prefixing skills.

Use it for:

• Build an agent that dynamically loads specialized research, coding, or analysis skills from a shared library without bloating the initial prompt.
• Compose agent capabilities from public skills repositories or custom skill packages with filtering and renaming.
• Create agents that execute bundled scripts (e.g., data processing, API calls) on demand as part of their task workflow.
• Manage a growing skill library across teams by loading skills from Git repositories or S3 with version control and composition.
• Teach an agent to handle domain-specific tasks by packaging instructions, reference documents, and helper scripts into a single skill folder.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Teaches agents to handle specialized tasks through modular skill packages that bundle instructions, reference documents, and executable scripts with progressive disclosure—skills load their full content only when needed.

Yes. The package is actively maintained, has low install friction, carries no known vulnerabilities, and fills a gap by supporting bundled resources and script execution. It is in alpha status but has been released since late 2025 and is in the top 5000 PyPI packages by downloads. Use it if you need agents to load specialized skills with reference materials or executable components.

Install

pydantic-ai-skills on PyPI

pip install pydantic-ai-skills

uv add pydantic-ai-skills

poetry add pydantic-ai-skills

Installing pydantic-ai-skills

Before you install

Active maintenance with a recent release (20 days old) and low install friction. Depends on three runtime packages: anyio, pydantic-ai-slim, and pyyaml. Supports Python 3.10 through 3.13.

License in practice

MIT license permits unrestricted use, modification, and distribution with minimal restrictions—suitable for commercial and proprietary projects.

Quickstart

from pydantic_ai_skills import SkillsCapability

agent = Agent(
    model='gateway/openai:gpt-5.2',
    instructions='You are a helpful research assistant.',
    capabilities=[SkillsCapability(directories=['./skills'])]
)
result = await agent.run('What are the last 3 papers on arXiv about machine learning?')

Requires Python 3.10 or later and an agent instance with a configured model.

Verify before relying

• Performance characteristics when managing large skill libraries or deeply nested skill hierarchies.
• Whether script execution sandboxing or permission controls are implemented for security.
• Specific version compatibility requirements with pydantic-ai-slim beyond the dependency declaration.

Package facts

Evidence: pydantic_ai_skills-1.3.0-py3-none-any.whl

Tags

More Python Modules packages

Further reading