claude-skillz

Python package for applying a compact, production-ready Copilot policy set to any repository.

Install

pip install claude-skillz

Quick Start

claude-skills init . --profile production-core
claude-skills verify . --profile production-core

This init command also scaffolds a PR workflow file at .github/workflows/claude-skills-verify.yml.

Import API

import claude_skills

report = claude_skills.install(".")
print(report.created_count)

What gets installed (production-core)

• 8 agent modes
• 5 instruction weights
• 11 high-value workflow skills
• 3 templates
• 1 PR CI workflow (.github/workflows/claude-skills-verify.yml)
• .github/copilot-instructions.md bootstrap with auto-dispatch rules

Why this model

• Small package footprint compared with full libraries
• Enforces behavior at repository level where Copilot actually reads policy files
• Keeps token/credit usage lower by bundling only core skills

Corporate and Restricted Networks

Use one of these patterns depending on your environment:

1) Private Index (recommended for enterprises)

Mirror claude-skillz to Artifactory/Nexus/Azure DevOps feed, then keep the same install command:

pip install claude-skillz

Set index config once on runners/developer machines:

pip config set global.index-url https://<private-index>/simple
pip config set global.trusted-host <private-index-host>

2) Proxy-Based Install

pip install claude-skillz --proxy http://<proxy-host>:<port>

3) Offline Wheel Install (air-gapped)

On a connected machine:

python -m pip wheel claude-skillz -w dist

On restricted machines:

pip install --no-index --find-links dist claude-skillz

CI for Every PR

The scaffolded workflow already runs:

claude-skills verify . --profile production-core

For non-GitHub CI, run the same command in your PR pipeline stage.

Detailed enterprise rollout options are documented in CORPORATE_INSTALL.md. Package release/mirroring steps are documented in PYPI_RELEASE.md.