skillsible 0.2.0

Ansible-style skill management for agent CLIs

skillsible

Ansible-style skill management for Codex, Claude Code, and agent CLIs.

What It Does

skillsible manages skills declaratively across machines and agents.

• define desired skills in a manifest
• target multiple agent CLIs
• preview changes with plan
• apply installs idempotently
• make workstation setup reproducible

MVP

The initial version focuses on skills.sh-style installs driven by a YAML manifest.

Example

version: 1

agents:
  - codex
  - claude-code

defaults:
  scope: global

skills:
  - source: obra/the-elements-of-style
    skill: writing-clearly-and-concisely

  - source: https://github.com/obra/the-elements-of-style
    skill: writing-clearly-and-concisely
    agents:
      - codex
    scope: project

CLI

uv run skillsible plan -f skills.yml
uv run skillsible apply -f skills.yml
uv run skillsible doctor

Install

From PyPI:

uv tool install skillsible
skillsible doctor

For a specific version:

uv tool install skillsible==0.2.0
skillsible doctor

From a checkout:

uv tool install .
skillsible doctor

From a built wheel:

uv build
uv tool install dist/skillsible-0.2.0-py3-none-any.whl
skillsible doctor

CI And Release

This repo includes GitHub Actions for:

• CI on pushes to main and pull requests
• building distributions with uv build
• publishing to PyPI when a tag like v0.2.0 is pushed

PyPI Trusted Publishing Setup

Before publishing will work, configure skillsible on PyPI to trust this repository's publish workflow:

• PyPI project: skillsible
• GitHub owner: srikalyan
• GitHub repository: skillsible
• Workflow file: .github/workflows/publish.yml
• Environment name: pypi

After that, pushing a tag such as v0.2.0 will trigger the publish workflow.

Workflow

The default branch is main and it is protected.

Expected flow:

1. Create a feature branch.
2. Open a pull request to main.
3. Let CI pass.
4. Merge to main.
5. Tag the merge commit with vX.Y.Z.
6. Push the tag to publish to PyPI.

See CONTRIBUTING.md for the release workflow in more detail.

Design Goals

• declarative desired state
• multi-agent support
• idempotent operations
• portable across machines
• extensible adapter model

Near-Term Roadmap

• manifest validation
• agent adapters
• drift detection
• lockfile support
• export current installed skills
• version pinning by tag or commit

Development

uv sync --dev
uv run pytest
uv run skillsible plan -f examples/skills.yml