Skill Registry Manager

Problem

When multiple agent skills exist in the same workspace, the issue is rarely that one skill is missing. It is that operators cannot quickly tell which skills are installed, whether their responsibilities overlap, and which skill should lead a task while others support it. skill-registry-manager treats installed skills as a manageable registry, making them visible, classifiable, and eligible for collaboration rules.

How It Works

The skill scans skills/**/SKILL.md and ignores non-business files such as .github/skills/**, *_GUIDE.md, *OVERVIEW*.md, and ordinary documentation. It prioritizes frontmatter fields like name and description, then uses the documented user requests, default workflow, and output requirements to map each skill to categories such as analysis, opportunity selection, risk control, workflow orchestration, skill governance, execution support, and delivery integration. Cross-category skills should declare primary_type and secondary_types.

Key steps include:
- Inventory installed skills: list name, path, description, and preliminary type.
- Build classification views: provide both per-skill and grouped-by-type perspectives.
- Define collaboration: assign primary_role, support_roles, and collaborates_with to each skill.
- Constrain collaboration: use one primary skill and at most two clearly value-adding support skills, rather than invoking everything at once.

Boundaries

It is suitable for skill cataloging, capability gap identification, duplicate responsibility review, and task collaboration recommendations. It does not replace business skills, and it should not assume that all skills should participate simply because they can collaborate. If the workspace contains few skills, the output should remain accurate and simple instead of inventing unnecessary structure.