Team Skill Libraries

A skill that lives on one developer's machine helps one developer. A skill that lives in a shared repository helps the entire team. Team skill libraries are the organizational infrastructure for distributing, discovering, and version-managing skills across engineering teams. They are the difference between "I built a cool skill" and "our team has a systematic approach to AI-assisted workflows."

The simplest implementation is a dedicated git repository — org-claude-skills — that contains skill directories organized by domain. Each skill directory is self-contained: front matter, core instructions, linked files, and a README documenting the skill's purpose, triggers, and usage examples. Developers install skills by copying the directory into their project's .claude/commands/ folder, or by symlinking to the shared repository.

Discovery is the primary challenge in team skill libraries. A library with fifty skills is useless if developers cannot find the skill they need. The solution is a manifest file at the repository root — a markdown index that lists every skill with its name, description, trigger phrases, and domain category. Developers search the manifest first, then install the relevant skills.

Version management follows semantic versioning. A skill's front matter includes a version field. When a skill is updated in the shared repository, the version increments. A CI check can compare installed skill versions against the latest versions in the shared repository and flag outdated installations. This is not automatic updates — automatic updates would break workflows that depend on specific skill behavior. It is visibility: "Your csv-normalizer skill is version 1.2. The latest version is 1.5. Changes in 1.3-1.5: added header detection, improved deduplication logic, fixed Unicode handling."