nyosegawa

Integrate, customize, debug, or upgrade Agent UI in external Codex App Server applications. Use when adding Agent UI, building a Codex UI, wiring a same-origin WebSocket bridge, composing React primitives, handling thread lifecycle, canonical resume, thread history, local media helper uploads, local desktop bridge policy, mobile drawer, overlay layers, host-gated workflow recipes, theming with --aui-* tokens, dynamic tools, Agent UI errors, or when the user says Agent UI導入, Codex UI, ローカルCodexアプリ, bridge, attachments, approvals, resume, local media, or upgrade Agent UI.

Plan Agent UI repository features before implementation. Use when the user asks to make a feature plan, plan before coding, research/plan/todo/goal prompt, solve a problem group, architecture plan, 実装前に計画, feature planning, PLAN/TODO作成, or goal-prompt generation for this repo. Do not use for small direct edits, review-only requests, npm release operations, upstream sync, browser QA, or external host-app integration unless the user explicitly asks for a planning package.

Operate Agent UI npm releases for @nyosegawa public packages with one reviewed release PR, local Changesets versioning, automatic main-push publishing, provenance, GitHub Releases, post-publish install smoke, rollback, and deprecation. Use when preparing or running an npm release, adding changesets, creating or reviewing release PRs, publishing packages, or debugging release workflow failures.

Review Agent UI repository changes, pull requests, branches, diffs, or Codex-generated updates for correctness, product boundary, protocol behavior, server bridge safety, package exports, tests, docs, UI regressions, and release risk. Use when the user asks for review, PR review, branch review, diff audit, safety check, or whether Agent UI changes are ready to merge.

Maintain Agent UI against upstream OpenAI Codex App Server schema drift. Use for protocol drift checks, schema refreshes, CODEX_PROTOCOL_COMMIT updates, method classification, and upstream sync PR creation.

Verify Agent UI browser-visible behavior with Playwright and agent-browser. Use for UI, CSS, layout, transcript, composer, approvals, usage, diagnostics, thread history, examples, mobile, overflow, focus, keyboard, click, hit-test, screenshot, accessibility snapshot, or real-local browser QA work.

Add, update, or review Agent UI examples, recipes, fixture routes, docs-site examples, Next.js/Vite integrations, local Codex web demos, theming recipes, uploads examples, dynamic tools examples, or example documentation. Use when working in examples/, docs/examples/, docs/recipes/, fixture routes, or browser-visible sample apps.

Validate Agent UI release readiness, package boundaries, build output, API snapshots, package resolution, Node compatibility, CI status, and publish risk. Use before release, before npm publish, after package export or declaration changes, when validating a PR for merge, or when GitHub Actions/package validation fails.

渡された文章を校正してブログに投稿する。Use when user asks to "記事を投稿", "ブログに投稿", "publish post", "記事を公開", "校正して投稿", "ブログ公開" etc.

セッションtranscriptを分析してドキュメントの効果を評価する。 doc参照がAgentの行動を改善したか、Context浪費か、腐敗して有害かを判定。 last-validated がないドキュメントにもgit履歴ベースで対応。 HTML reportでper-doc健全性、Context budget、freshness、推奨アクションを表示。 Use when user says "audit docs", "docs audit", "run docs-auditor", "ドキュメント監査", "doc ROI", "ドキュメントの効果分析".

Gemini画像生成APIを使ってブログ記事やドキュメント用の図・イラストを生成する。Use when user asks to "画像生成", "図を作って", "generate image", "create diagram", "イラスト作って", "図示して", or wants to create visual assets for articles.

MCPサーバーの「Light版」を生成する。descriptionを1行に圧縮し、ベストプラクティスをAgent Skillとして分離する。Use when user asks to "MCP Light版を作って", "MCPを軽量化", "Light MCP", "mcp-light", "create light mcp", "compress mcp descriptions", "MCP description圧縮", or wants to reduce MCP tool definition token usage.

Build high-quality Remotion promo and intro videos for any app type (web, mobile, API/SDK, developer tool, AI product) using app-type blueprints, timeline patterns, and frame-capture QA. Use when users ask to create a product promo, teaser, app intro, UX flow video, or Remotion-based demo video.

Analyzes Claude Code session transcripts to evaluate skill portfolio health — routing errors, attention competition between descriptions, and coverage gaps. Generates an interactive HTML report with per-skill health cards, competition matrix, attention budget analysis, and actionable patches. Unlike skill-creator which optimizes individual skills in isolation, skill-auditor optimizes the portfolio as a system, detecting cross-skill attention theft and cascade risks. Use when user says "audit my skills", "skill audit", "run skill-auditor", "analyze skill routing", "check skill competition", "portfolio health", "スキル監査", "スキルの精度を分析", "スキルルーティング分析".

Initialize Unity project and configure URP/Post Processing. Use when starting a new project, setting up URP pipeline, configuring post processing, or creating folder structure.

Diagnose and fix Unity C# compilation errors. Use when compilation fails, CS error codes appear, Unity_GetConsoleLogs shows errors, or user asks to fix compile errors.

Manage Unity scenes, GameObjects, and components through Unity MCP. Use when building scenes, placing objects, setting up hierarchy, creating materials, or when user asks to construct or modify a scene.

Unity C# coding standards and best practices reference. Use when writing MonoBehaviour scripts, creating C# classes, reviewing Unity code, or when user asks about naming conventions, component patterns, or code quality.

Run and analyze Unity tests via Unity MCP. Use when running EditMode or PlayMode tests, creating test scripts, analyzing test failures, or when user asks to test code.

Generate HTML reports and structured metrics from model evaluation results. Creates publication-ready reports with embedded outputs (images, audio, charts) and metrics.json for cross-model comparison. Use when generating reports, writing metrics, creating evaluation summaries, or formatting benchmark results. Triggers on "generate report", "write metrics", "create report", "evaluation summary", "benchmark results", "format results".

Execute model inference on GPU cloud providers. Handles code generation, deployment, execution, and result collection across HF Inference API/Endpoints, Colab, Modal, beam.cloud, Vast.ai, and RunPod. Use when running models on GPU, deploying to cloud, executing notebooks, or troubleshooting GPU execution failures. Triggers on "run on GPU", "execute model", "deploy to modal", "colab notebook", "beam deploy", "HF inference", "HF endpoints", "vast", "runpod".

Investigate model specifications, requirements, and evaluation strategy. Use when researching a model before benchmarking: reading HuggingFace model cards, estimating VRAM requirements, selecting GPU providers, and determining evaluation approach. Triggers on "model research", "investigate model", "model info", "VRAM estimate", "which provider", "model card".

Autonomous model validation and benchmarking. Investigates any ML model (LLM, image gen, TTS, time series, etc.), runs it on GPU cloud, evaluates quality and performance, and generates HTML reports. Use when user asks to verify, benchmark, evaluate, or test a model. Triggers on "verify model", "benchmark", "evaluate model", "test model", "run benchmark", "model evaluation", "モデルを検証", "ベンチマーク", "モデルを試して".

活字ビジネス文書（請求書・領収書・名刺）の合成データを生成する。 LLM の創造性でリアルな日本語ビジネスデータとユニークなレイアウト (HTML/CSS) を生成し、 Playwright で画像化する。 Use when: "データ生成", "generate-business-doc", "合成データ", "ビジネス文書生成", "請求書生成", "領収書生成", "名刺生成", "structured_eval データ"

Modal上でOCRモデルを動かすスクリプトの新規作成・デバッグ・動作確認を行う。 このプロジェクトのmodal_scripts/パターンに準拠したスクリプトを生成し、 既知のハマりどころを事前回避する。 Use when: "Modal", "modal_scripts", "Modalスクリプト", "Modalで動かす", "GPUで実行", "新しいOCRモデルを追加", "Modalモデル追加", "modal run".

agent-browser でアクセシビリティツリーを取得し、UI 構造を機械的に検証する。 CLS 測定、アニメーション完了待ち、スナップショット比較。 Use when: "UI確認", "見た目チェック", "コンポーネント検証", "visual check", "accessibility tree", "CLS", "アクセシビリティ", "スナップショット".

Operate Notion workspaces via ncli (MCP + REST API). Covers page search/create/update, database create/query, view management, comments, file upload, and direct REST API access. Use when user asks to "Notion に書いて", "ページ作って", "タスク管理", "DB 作成", "Notion で検索", "議事録", "ファイルアップロード", "create a Notion page", "track tasks in Notion", "upload file to Notion", or any Notion workspace operation. Also triggers on "notion" keyword in requests.