Kurogoma4D

方針A（反復セルフレビュー）で 1 つの GitHub issue を 1 本の PR に実装する。 独立 Subagent(github-issue-implementer)で実装後、独立 Subagent(code-reviewer)による セルフレビューと修正を指摘点がなくなる（LGTM）まで繰り返し、PR をオープンのまま残す（マージしない）。 各マイルストーンを PR の行動記録(runlog)に残し、速度・反復回数はその記録から集計する。 `/pr-strategy-iterate <issue-number>` で起動する。

方針B（分割 PR）で 1 つの GitHub issue を、ある程度細かい単位の複数 PR に分割して実装する。 各 PR を独立 Subagent(github-issue-implementer)で実装し、独立 Subagent(code-reviewer)による セルフレビューと修正を「1 ターンのみ」行う（再レビューしない）。全 PR をオープンのまま残す。 各マイルストーンを PR の行動記録(runlog)に残し、速度・分割数はその記録から集計する。 `/pr-strategy-split <issue-number>` で起動する。

PR 戦略比較実験の「品質」を測るスキル。1 つの issue に対する最終的な差分（方針A は 1 本の PR、 方針B は複数 PR の和集合）を、横断文脈を持たない独立 Subagent に採点させる。 オーケストレーターは自分では採点せず、Subagent が返した構造化スコアをそのまま quality.json に記録する。 あわせて PR の行動記録(runlog)を辿って速度・反復回数を集計し metrics に反映する。 `/pr-quality-judge <issue-number> <pr-number>[,<pr-number>...]` で起動する。

Automatically consume open GitHub issues one by one. Fetches the oldest open issue, delegates implementation to the issue-implementer agent, runs code review via the code-reviewer agent, iterates on feedback, merges the PR, and moves on to the next issue. Invoke with `/auto-issue-worker`.

Run the auto-issue-worker flow against a single GitHub issue: implement it via the issue-implementer agent, review the PR once with the code-reviewer agent, apply one round of fixes for any findings, then stop with the PR left open (no merge). Use when you want a single issue taken from open to a ready-for-review PR without merging. Invoke with `/issue-worker <issue-number>`.

Run an authorized penetration test against the project after some implementation exists: combine static code analysis and dynamic testing against the running app to detect vulnerabilities, then register each confirmed finding as a GitHub issue. Use once enough of the project is implemented to be worth probing for security flaws. Invoke with `/pentest`. The created issues are consumed by `/auto-issue-worker`.

Interview the user to gather project requirements and consolidate them into a structured `spec.md` at the repository root. Use when starting a new project or feature and the requirements are not yet written down. Invoke with `/spec-builder`. Produces the input consumed by the `spec-to-issues` skill.

Read `spec.md`, break it down into discrete implementation tasks, and register them as GitHub issues in the project repository. Use after a spec exists (e.g. produced by the `spec-builder` skill) and the user wants to turn it into actionable issues. Invoke with `/spec-to-issues`. The created issues are consumed by `/auto-issue-worker`.

Automatically consume open GitHub issues, in parallel where dependencies allow. The main agent acts as Project Manager: it builds a dependency-aware work plan, dispatches issue-implementer agents (Tech Specialists) concurrently, reviews each PR with a panel of specialist reviewers, consolidates findings, iterates on fixes, and merges. Repeats until no open issues remain. Invoke with `/auto-issue-worker`.

Run the auto-issue-worker flow against a single GitHub issue: implement it via the issue-implementer agent (Tech Specialist), review the PR once with a panel of specialist reviewers, consolidate the findings, apply one round of fixes, then stop with the PR left open (no merge). Use when you want a single issue taken from open to a ready-for-review PR without merging. Invoke with `/issue-worker <issue-number>`.

Run the full kickstarter workflow — setup, spec, issues — in one command.

Fill in the template placeholders across the `.claude/` agents and skills based on an interview about the target project. Use right after copying this kickstarter template into a project, or when project details change and the placeholders need refreshing. Invoke with `/template-setup`.

Pull updates from the kickstarter template into a working repository created from it, using a per-file 3-way merge so project-specific placeholder values are preserved and only the template's own changes are applied. Use when the upstream template has been updated and you want those changes reflected here. Invoke with `/template-update`.

Interview the user to gather project requirements and consolidate them into a structured `spec.md` at the repository root. Use when starting a new project or feature and the requirements are not yet written down. Invoke with `/spec-builder`. Produces the input consumed by the `spec-to-issues` skill.

Harden the project against software supply-chain attacks across two layers: project dependencies (lockfile pinning, audits, install-time script controls) and the CI/build pipeline (SHA-pinned actions, least-privilege tokens, Dependabot/Renovate). Sets up the guardrails with the user's approval, then audits existing dependencies and files a GitHub issue per remaining risk. Use during environment setup, or whenever dependencies or CI change. Invoke with `/supply-chain-guard`. The created issues are consumed by `/auto-issue-worker`.

Interview the user about a single ad-hoc task, then register it as one GitHub issue in the project repository. Use when an individual task comes up (a bug, a small feature, a chore) and the requirements are not yet written down — without going through `spec.md`. Invoke with `/task-to-issue`. The created issue is consumed by `/auto-issue-worker`.

Automatically consume open GitHub issues one by one. Fetches the oldest open issue, delegates implementation to the issue-implementer agent, runs code review via the code-reviewer agent, iterates on feedback, merges the PR, and moves on to the next issue. Invoke with `/auto-issue-worker`.

tsumugiをワンショットモードで実行し、動作確認を行う。 --event-log でイベントを記録し、stdout出力とログの両方から動作を検証する。 `/run-oneshot` で起動。引数にプロンプト文字列を指定可能。

tsumugiをTUIモードで起動し、対話的な動作確認を行う。 --event-log でイベントログを出力し、Claudeがログファイルを読んで動作を検証する。 `/run-tui` で起動。引数にテストしたいプロンプトを指定可能。

tsumugiプロジェクトの動作確認を行う。ビルド、lint、フォーマット、テスト、 バイナリのスモークテストを順に実行し、結果をレポートする。 `/verify` で起動。

Automatically consume open GitHub issues one by one. Fetches the oldest open issue, delegates implementation to the issue-implementer agent, runs code review via the code-reviewer agent, iterates on feedback, merges the PR, and moves on to the next issue. Invoke with `/auto-issue-worker`.

Build project-level design systems based on Material Design principles. Use when creating a new design system, extracting one from an existing product, extending tokens or components, or reviewing UI for design system compliance. Outputs a project-level Agent Skill with token definitions and component specifications.

Check Pull Request on current branch, and obtain unresolved review comments. `gh` command is required. Use this when user instructed with `PR review comment` `PR レビューコメント` etc.

ShiroGuessr "Shiro Gallery" UI design guideline. Use when implementing or modifying UI for ShiroGuessr on either Android (Jetpack Compose) or iOS (SwiftUI). Defines design concept, color system, typography, screen-by-screen guidelines, animations, and component style guide. All UI changes MUST follow these guidelines.

Reference the Kurogoma4D/ShiroGuessr-iOS repository using gh CLI commands. Use when the user mentions "iOS版" (iOS version), "現行プロジェクト" (current project), or needs to reference existing implementations from the iOS version of ShiroGuessr. Supports code search, keyword search, and file content retrieval.

Automatically consume open GitHub issues one by one. Fetches the oldest open issue, delegates implementation to the issue-implementer agent, runs code review via the code-reviewer agent, iterates on feedback, merges the PR, and moves on to the next issue. Invoke with `/auto-issue-worker`.

Automatically consume open GitHub issues one by one. Fetches the oldest open issue, delegates implementation to the issue-implementer agent, runs code review via the code-reviewer agent, iterates on feedback, merges the PR, and moves on to the next issue. Invoke with `/auto-issue-worker`.

Reference the Kurogoma4D/shiro-guessr repository using gh CLI commands. Use when the user mentions "Web版" (Web version), "現行プロジェクト" (current project), or needs to reference existing implementations from the web version of ShiroGuessr. Supports code search, keyword search, and file content retrieval.

Automatically consume open GitHub issues one by one. Fetches the oldest open issue, delegates implementation to the issue-implementer agent, runs code review via the code-reviewer agent, iterates on feedback, merges the PR, and moves on to the next issue. Invoke with `/auto-issue-worker`.