[Feature]: Support skill.json in source repos for faster discovery

[xtfer]

Problem

When npx skillman add vercel-labs/skills, skillman clones the repo and delegates to the skills CLI to discover available skills via SKILL.md scanning. There's no way for a skill author to declare a structured index of what's in their repo — consumers have to scan the filesystem every time.

Proposed Solution

Read an optional skill.json (singular) from the root of source repos as a fast-path discovery mechanism. This is a publisher-side file with different semantics to skillman's own consumer-side skills.json (plural).

When a source repo contains skill.json:

• npx skillman add owner/repo can list available skills without scanning
• Skill names in the consumer manifest can be validated against the source
• Future: integrity hashes enable cache-based deduplication

Example skill.json:

{
  "name": "my-skills",
  "version": "1.0.0",
  "description": "A collection of useful skills",
  "skills": [
    { "name": "skill-a", "path": "./skill-a", "description": "..." },
    { "name": "skill-b", "path": "./skill-b", "description": "..." }
  ]
}

This could be consumed directly by skillman, or inherited for free if the upstream skills CLI (vercel-labs/skills) adds support — since skillman already delegates to it.

Alternatively, support the linked issue on vercel-labs/skills.

No conflict with skills.json

skill.json (singular, publisher-side, in skill repos) vs skills.json (plural, consumer-side, in project root). Different file, different location.

Spec

Full schema and specification: https://github.com/velvet-tiger/skill.json

Related

• [Proposal]: Add skill.json as an optional package-level metadata file in https://github.com/agentskills/agentskills
• [Feature]: Support skill.json for publisher-side skill discovery in https://github.com/vercel-labs/skills