Installation

okf-schema is published to PyPI and can be installed with any Python package manager.

With pip

(it is not recommended to install a CLI, uv is always to be preferred when available)

pip install okf-schema

Verify the installation

okf-schema --version

Upgrade (uv)

uv tool install --upgrade okf-schema

Using OKF-Schema as a library

Install as a Python library for programmatic use:

uv add okf-schema

Install the agent skills

The skills are included in the package and installed by the command that owns each skill family. Run the commands for the families you want to use:

okf-schema install-skills
okfkb install-skills
okfreq install-skills

By default, skills are installed globally under ~/.copilot/skills. To install them in the current project instead, choose the directory supported by your agent:

# GitHub Copilot project skills
okf-schema install-skills --local-copilot
okfkb install-skills --local-copilot
okfreq install-skills --local-copilot

# Agent Skills project directory
okf-schema install-skills --local-agents
okfkb install-skills --local-agents
okfreq install-skills --local-agents

An explicit destination can be used when the project has its own skill directory:

okf-schema install-skills ./vendor/agent-skills

The commands install these families:

  • okf-schema covers generic bundle authoring and validation.

  • okfkb includes knowledge-base authoring, findings, and consolidation skills.

  • okfreq includes requirements authoring and requirements-maintenance skills.

Invoke a skill by naming it in your request. For example:

Use okfkb-record-findings to record this observation in the current knowledge base: exporting 10,000 rows took 4.8 seconds with version 1.4 and the sample customer data set.

The agent uses the skill to gather context, edit the correct files, and run the project-prescribed checks. Review its changes before committing them. The CLI remains available for deterministic operations such as validate, lint, and update.

Next steps