nwiizo

Rust/Axum などで SPA 向け OAuth 2.0 / OpenID Connect BFF を実装・レビュー・検証する。ブラウザにアクセストークンやリフレッシュトークンを出さない設計、HttpOnly Cookie セッション、CSRF/CORS/Origin 検証、Hydra などの認可サーバー連携、Playwright でのログイン往復検証を扱うときに使用。

公式翻訳との比較から導出した技術書翻訳の品質基準。翻訳作業時に自動参照。

Build→Test→Feedback→Refineの反復改善ワークフロー。ツール開発時のフィードバックループとOSS検証を支援。

Claude Code TaskList と vibe-ticket の同期。タスク管理の状態を双方向で同期する。

LLM コーディングの 4 大失敗パターン（暗黙の仮定／過剰実装／関係ない箇所への手出し／成功基準の曖昧さ）を抑制するための行動規範。コード生成・編集・リファクタ時に参照。

AWS アカウントの FinOps 調査・コスト削減分析。ReadOnly + MFA 環境での非対話認証突破、Cost Explorer の読み解き罠、RI/SP の誤認パターン、レポート数値の分母混在を避けるための学び集。

Spawn a parallel coding-agent session in its own jj workspace. Uses `jj_agent` if the nwiizo/jujutsu.fish plugin is installed, otherwise falls back to raw `jj workspace add`. Use when starting a second (or Nth) agent on a different task without disturbing the current working copy. Invoke manually; creates a workspace directory on disk.

In any jj (Jujutsu) repository, complete the "commit" pattern — describe the current working copy with a message, then open a fresh change for the next unit of work. Use when a coherent edit is finished and should be preserved before moving on. Invoke manually; mutates state.

README、AGENTS、CLAUDE、ツール別READMEを俯瞰し、重複・古い前提・読みにくい構成・不要ファイルを整理する。ドキュメント整備やリポジトリを読みやすくしたいときに使用。

AIエージェント履歴を安全に集計し、再利用できる skills / agents / rules / scripts 候補だけを抽出する。履歴本文を公開せず、秘密情報を先に検出して抽象化したいときに使用。

jj リポジトリで検証後に現在の変更を describe し、main bookmark を進め、origin/main へ push する。ユーザーが「コミットしてpush」と依頼したときに使用。

このスキルは、ユーザーが「プロンプトをレビューして」「対話履歴を分析して」「理解度を診断して」 と依頼したとき、または /prompt-review で呼び出されたときに使用する。 過去のAIエージェント対話履歴（Claude Code, GitHub Copilot Chat, Cline, Roo Code, Windsurf, Antigravity, OpenCode）を読み取り、技術理解度・プロンプティングパターン・AI依存度を推定してレポートを生成する。

Add config to this macOS dotfiles repo. Use for env vars, fish abbreviations/functions/plugins, direct symlinked config files, tool integrations, shared agent config, and app config such as Neovim, Ghostty, Warp, Git, Bat, Atuin, tealdeer, and gh.

Add a package to this macOS dotfiles repo. Use when the user asks to install, add, or migrate a CLI package or GUI app, for example "add jq", "ripgrep を入れて", "lazygit 入れたい", "brew install foo を repo 管理に", "kubectl 追加して". Homebrew owns binaries here; update Brewfile, validate with brew bundle, apply with brew bundle, and verify the executable.

AWS アカウントの FinOps 調査・コスト削減分析。ReadOnly + MFA 環境での非対話認証突破、Cost Explorer の読み解き罠、RI/SP の誤認パターン、レポート数値の分母混在を避けるための学び集。

agent 向けテキスト指示（skill / slash command / task プロンプト / CLAUDE.md 節 / コード生成プロンプト）を、バイアスを排した実行者に動かしてもらい、両面（実行者の自己申告 + 指示側メトリクス）で評価して反復改善する手法。改善が頭打ちになるまで回す。プロンプトや skill を新規作成・大幅改訂した直後、またはエージェントの挙動が期待通りにならない原因を指示側の曖昧さに求めたいときに使う。

Duplicate code detection using similarity-rs. Identifies refactoring candidates based on semantic similarity.

Duplicate code detection using similarity-rs. Identifies refactoring candidates based on semantic similarity.

Run performance benchmarks with criterion. Compare against baselines and profile hotspots.

Implementation verification - runs format, lint, test, and build checks on the workspace.

Production readiness audit - checks unwrap elimination, error handling, clippy, docs, channel usage, and test count.

Release deployment process for ccswarm. Version update, quality gates, build, tag, publish.

Parallel development workflow using git worktrees. Each agent gets an isolated worktree for safe concurrent work.

Human-in-the-loop approval workflow for high-risk agent operations (file deletions, deployments, config changes).

Khononov's Balanced Coupling model reference. Auto-loaded when analyzing coupling patterns, reviewing architecture, or interpreting analysis results.

Run coupling analysis and interpret results. Use when analyzing Rust project coupling patterns and generating improvement proposals.

Quick health check for coupling balance score. Use for daily score verification and CI quality gates.

Modularity review using Balanced Coupling model. Combines automated analysis with semantic code review to find implicit coupling and design issues.

Start interactive web UI for coupling analysis visualization. Graph exploration with filters, hotspots, and export.

Run mutation testing with cargo-mutants to evaluate test quality. Identifies untested code paths.

Explain coupling issue types in detail with detection conditions, code examples, and solution approaches.

Identify high-priority refactoring targets using hotspot scoring. Ranks modules by issue severity and coupling count.

Run full quality gate before release — fmt, clippy, test, mutants, similarity, coupling.

Run cargo-coupling and similarity-rs to identify refactoring targets, then create a prioritized plan.

Systematic bug investigation and fix workflow. Use when fixing bugs, investigating issues, or debugging test failures.

Pre-release verification workflow - version, security audit, performance, cross-platform builds, documentation.

Development environment setup - Rust toolchain, cross-compilation targets, dependencies, git hooks, IDE config.

Split complex tasks into sequential steps with parallel subtasks. Use for multi-step workflows needing dependency ordering and adaptive planning.

Comprehensive quality assurance - tests, linting, formatting, code analysis, and release build verification.

Systematic code refactoring - analysis, pattern detection, incremental changes with verification. Use for code improvement and tech debt reduction.

Test and deepen the load-bearing claims in a Marp presentation. Use when punchlines feel generic, causal claims are unsupported, conclusions overreach their evidence, trade-offs or boundary conditions are missing, or the audience cannot tell how an insight should change a decision.

Validate and prepare Marp slides for public release. Use before publishing or pushing changes that affect slide Markdown, themes, images, PDFs, README links, brand packs, or build configuration, especially when clean-clone reproducibility and source-to-artifact consistency matter.

Review a Marp presentation’s logical flow from the audience’s point of view. Use when checking section order, slide-to-slide reasoning, unanswered questions, hidden prerequisites, promise fulfillment, transitions, or whether a proposed reordering will make a talk easier to follow.

Review the narrative and audience-empathy design of a Marp presentation. Use when a technically correct talk feels flat, the opening lacks stakes, examples do not connect to the audience, tension or agency disappears, the ending lacks payoff, or proposed storytelling risks inventing personal experiences.

Coordinate and synthesize multiple specialist reviews of a Marp presentation into one coherent edit strategy. Use when a deck needs flow, claim depth, narrative, redundancy, and wit reviews together, or when specialist recommendations conflict over adding, cutting, moving, or rewriting slides.

Review humor, satire, irony, metaphor, and rhetorical wit in a Marp presentation for purpose, precision, speaker voice, audience fit, and context-collapse safety. Use when a talk contains jokes or playful phrasing, needs sharper memorable language, risks punching down or being misread in screenshots, or should remain serious without forced humor.

Find and safely remove redundancy in a Marp presentation while preserving comprehension, emphasis, pacing, and accessibility. Use when a deck is too long, repeats claims or examples, contains thin or overloaded slides, needs a target duration, or requires evidence-based cut, merge, compress, and keep decisions.

スライドビルド - Marpスライドを `--no-stdin` 付きで HTML/PDF 出力する

スライドレビュー - Marpプレゼンテーションを聴衆起点・構成・容量・配色・視覚装飾・ファクト・表現の7観点で総合評価

構成チェック - プレゼンテーションの骨格（全体構成・目次整合・Why→What→How順序・繋ぎスライド）を評価する

ファクトチェック - スライド内の数値・引用・技術仕様・人名・年号の事実関係をWebSearch/WebFetchで検証

Spawn a parallel coding-agent session in its own jj workspace using `jj_agent`. Use when starting a second (or Nth) agent on a different task without disturbing the current working copy. Invoke manually; creates a workspace directory on disk.

In a jj repository, complete the "commit" pattern — describe the current working copy with a message, then open a fresh change for the next unit of work. Use when a coherent edit is finished and should be preserved before moving on. Invoke manually; mutates state.

Add a new jj abbreviation across the 4 required sites (register, erase, README, test) with verification. Invoke manually; has side effects.

Scaffold a new jj_fzf_* picker following the shared contract — dependency guards, cancel code 130, tab-delimited templates, binding wiring, README entry. Invoke manually; has side effects.

Generate Japanese daily reports, reflection prompts, guides, reviews, and trend reports from Claude Code or Codex work logs. Use when the user asks for nippo, 日報, daily, reflection, guide, report, review, insight, trend, or wants to summarize recent Claude Code/Codex work.

Claude Code / Codex のセッションログから日報・リフレクション・インサイトを生成する。 /nippo と /nippo daily で日報、/nippo reflection で内省の問い、/nippo guide で学習支援、 /nippo report で進捗報告、/nippo review で自己評価、/nippo insight で深い振り返り、 /nippo trend で長期変化分析を生成する。Rust バイナリ (nippo) でデータを収集する。