Add capability declaration fields to SKILL.md frontmatter (tools, mcp-servers, hooks, model)
- 主要言語
- Shell
- スター
- 11.2k
- フォーク
- 1.9k
- 平均マージ
- 14時間 16分
- マージ済み PR(30日)
- 6
説明
## Brief
**Canonical issue:** https://github.com/microsoft/vscode-copilot-chat/issues/5101
Cross-filed at the GitHub Copilot CLI repo because SKILL.md is a cross-tool format consumed by both VS Code Copilot Chat and Copilot CLI plugin install. The gap and the proposal apply identically here.
## Summary (1 paragraph)
Custom agents (`.agent.md`) declare their capabilities in frontmatter — `tools`, `mcp-servers`, `hooks`, `model`. Skills (`SKILL.md`) cannot — they currently support only `name`, `description`, `argument-hint`, `user-invocable`, `disable-model-invocation`. The asymmetry forces every consuming agent to know and re-declare a skill's capability requirements, leading to silent under-capability and ecosystem-specific workarounds. Proposed: extend SKILL.md frontmatter with the same optional fields, with skill-declared `tools`/`mcp-servers` unioning with consuming agents, skill-declared `hooks` activating with the skill, and `model` as advisory.
## Why filed here
`copilot plugin install` loads SKILL.md files; the same capability-declaration gap applies to the CLI surface. Closing it natively at the SKILL.md level reduces the need for plugin-level bundling workarounds (whole plugin enables together, coarser than per-skill) and aligns skill composability across both surfaces.
See the canonical issue for full proposal, use cases, inheritance semantics, and references.
コントリビューションガイド
調査の方向性
Start with the canonical issue at microsoft/vscode-copilot-chat#5101 and trace where copilot plugin install loads SKILL.md files. Determine the relevant frontmatter parsing and capability-inheritance entry points, then validate that the proposed fields and semantics work across the CLI and VS Code surfaces with appropriate tests.
索引モデルが issue の本文から書いたものです。
評価
- 技術スタック
- github
- 領域
- cli, tooling
- issue の種類
- 機能追加
- 難易度
- 5/5
- 見積もり時間
- 1週間以上
- 活発さ
- 静か
- 明瞭さ
- 説明が足りない
- 初心者へのやさしさ
- 35/100