[Feature]: Skill resources for presets (copy-mode aware composition and preset-owned resources)

オープン
  • 何か問題がありますか？
#5,382 コメント 0 件 リアクション 0 件 担当者 0 名 GitHub で見る

メンテナーはふだん 1 日以内に返信

まだ誰も着手していません。

評価

難易度
5/5
見積もり時間
1週間以上
初心者へのやさしさ
35/100
issue の種類
機能追加
明瞭さ
明確に書かれている
活発さ
活発
技術スタック
python
領域
tooling

調査の方向性

Start with presets/_resolver.py at PresetResolver.resolve_content and compare its path rewriting with the copy-mode handling from #5366. Then read extensions/_skill_resources.py and the extension resource lifecycle tests to trace validation, copying, registry ownership, and removal. Done means the acceptance criteria define stacked resource ownership and preset removal behavior, with matching tests for composition and resource lifecycle.

索引モデルが issue の本文から書いたものです。

説明

triage-can-wait

Problem Statement

#5366 lets extension commands declare resources that are copied into their generated agent skills ("copy mode"; skill bodies are not path-rewritten). Presets are not covered:

  1. Composition: when a preset overrides a copy-mode extension command with append/prepend/wrap, PresetResolver.resolve_content (presets/_resolver.py ~l.789) still applies rewrite_extension_paths to the extension base layer. The composed skill therefore references .specify/extensions/<id>/... instead of the copied skill-relative resources. It still resolves, but is inconsistent with copy mode.
  2. Preset-owned resources: presets that override or provide skills cannot bundle their own resources (references, templates, scripts) into the skill directory.

Proposed Solution

  • Make preset composition copy-mode aware for skill outputs: when the winning base layer is a copy-mode extension command and the output is a skill, do not rewrite its paths. Command/template consumers keep today's rewriting.
  • Add an optional resources list to preset command/template entries in preset.yml (same pattern dialect as extension resources), copied into the preset's skill output with hash-tracked ownership in the preset registry, reusing the extension resource helper (extensions/_skill_resources.py).

Alternatives Considered

  • Keep presets on extension-local paths only (today's behaviour): works, but preset authors cannot ship skill-relative files.
  • Only fix composition (item 1) and defer preset-owned resources.

Component

Presets / skills rendering

AI Agent (if applicable)

All SKILL.md-based integrations

Use Cases

  • A team preset appends review rules to an extension skill and ships its own references/team-rules.md.
  • A preset wraps a copy-mode extension command and the composed skill should keep skill-relative references.

Acceptance Criteria

  • Composed skill bodies over a copy-mode extension base are not path-rewritten; non-skill outputs unchanged.
  • Preset resources are validated like extension resources and copied as regular files.
  • Ownership defined for path conflicts between an extension and one or more presets writing the same skill directory (who wins, who owns, what happens on preset removal and priority changes).
  • Defined behaviour for stacked layers (append/prepend/wrap): whose resources are copied.
  • On preset remove, preset-owned resources are removed and the extension's resources/skill are restored.
  • Tests mirror the extension resource lifecycle tests.

Additional Context

Deferred out of #5366 by design; that change only restores extension copy-mode skills (with resources) after preset remove.

AI Disclosure

Drafted with the pi coding agent (orchestrator model: Claude Opus 5.5 via GitHub Copilot; implementation/review subagents: GPT-6.1-terra, GPT-6-astra), human-supervised by the filer. Reproductions were run locally by the agent and reviewed by the filer before filing.

主要言語
Python
スター
149k
フォーク
13.4k
平均マージ
1日 5時間
マージ済み PR（30日）
223

環境構築

Codespaces で開く

このプロジェクトの開発コンテナを、あなたの GitHub アカウントでブラウザ上に起動します。

はじめの一歩

  1. issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
  2. 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
  3. リポジトリをフォークし、ブランチを切って変更します。
  4. issue 番号を参照したプルリクエストを送ります。

github/spec-kit のほかの issue

github/spec-kit の issue をすべて見る

似ている issue

Python の issue をもっと見る

新しい issue をメールで受け取る

初心者向けの GitHub issue を短くまとめたダイジェスト。