Feature request: standardize skill metadata (origin, version, license)

Problem

Claude Code skills currently have minimal metadata — just name and description in YAML frontmatter. As the skill ecosystem grows (ECC, community repos, auto-extracted, user-created), users accumulate dozens of skills with no way to identify their provenance, version, or license.

Real-world impact

After 3 months of use, my ~/.claude/skills/ directory contains 66 files from at least 5 different sources:

• Everything Claude Code (ECC) — 28 skills
• Community repositories — 1 skill
• Self-authored — 1 skill
• Auto-extracted by continuous-learning — 28 skills
• Generated by skill-create — 1 skill

Without opening each file and comparing against source repositories, there is no way to tell them apart. This makes updates, cleanup, and sharing impractical at scale.

Proposal

Standardize optional metadata fields in skill YAML frontmatter:

---
name: coding-standards
description: Universal coding standards...
origin: ECC                    # Where this skill came from
version: 1.2.0                 # Semantic version (optional)
license: MIT                   # License identifier (optional)
---

origin field (highest priority)

Indicates where the skill was installed from. Suggested values:

| Value | Meaning |
|-------|---------|
| original | User-created |
| {org/repo} | From a specific GitHub repository |
| auto-extracted | Generated by continuous-learning or similar |

Why this matters for the ecosystem

• npm has package.json with author, license, repository fields
• Python has pyproject.toml with project metadata
• VS Code extensions have package.json with publisher info

Claude Code skills are the equivalent of packages/extensions but currently lack any provenance metadata. As skill sharing grows (marketplaces, team sharing, community repos), this gap will become more painful.

Current workaround

I created a convention using either YAML frontmatter fields or HTML comments:

origin: ECC           # inside frontmatter

<!-- origin: original -->    <!-- before content -->

And wrote a script to bulk-tag all existing skills. This works but is fragile and non-standard.

Related

• Filed affaan-m/everything-claude-code#246 requesting ECC add origin: ECC to distributed skills
• This is a broader ecosystem request for Claude Code to recognize and potentially validate these fields