Harness engineering beginner tutorial, from 0 to 1
TypeScript agentagenticagentic-aiaiai-agentai-agentsdshdsh-pluginharnessharness-engineeringharness-frameworkllm
Sync translated skill docs
sanbuphyf8f9c13d173d
diff --git a/docs/ar/skills/index.md b/docs/ar/skills/index.mdindex 1671992..7b1529b 100644--- a/docs/ar/skills/index.md+++ b/docs/ar/skills/index.md@@ -25,11 +25,12 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator### أنماط مرجعية-تتضمن المهارة 6 وثائق مرجعية مفصلة:+تتضمن المهارة 7 وثائق مرجعية مركزة:| النمط | متى تستخدمه ||---------|-------------|| Memory Persistence | عندما ينسى الوكيل بين الجلسات |+| Skill Runtime | تغليف workflows قابلة لإعادة الاستخدام كمهارات || Context Engineering | إدارة ميزانية السياق والتحميل عند الحاجة || Tool Registry | أمان الأدوات والتحكم في التوازي || Multi-Agent Coordination | التوازي و workflows التخصص |@@ -44,7 +45,11 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator- `feature-list.json` — JSON Schema ومثال قائمة ميزات- `init.sh` — سكربت تهيئة قياسي- `progress.md` — قالب سجل تقدم الجلسة-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — قالب تسليم الجلسة++### Scripts++تتضمن المهارة أيضا scripts نقية بـ Node.js لإنشاء scaffold والتحقق وتقارير HTML و benchmark بنيوي.### كيف بُنيت هذه المهارةdiff --git a/docs/de/skills/index.md b/docs/de/skills/index.mdindex 4a4cf1f..7136e61 100644--- a/docs/de/skills/index.md+++ b/docs/de/skills/index.md@@ -25,11 +25,12 @@ Um ihn mit Claude Code zu verwenden, kopiere das Verzeichnis `harness-creator/`### Referenzmuster-Der Skill enthält 6 ausführliche Referenzdokumente:+Der Skill enthält 7 fokussierte Referenzdokumente:| Muster | Wann verwenden ||---------|----------------|| Memory Persistence | Der Agent vergisst zwischen Sessions |+| Skill Runtime | Wiederverwendbare Workflows als Skills paketieren || Context Engineering | Kontextbudget, JIT-Laden || Tool Registry | Tool-Sicherheit, Nebenläufigkeitskontrolle || Multi-Agent Coordination | Parallelität, Spezialisierungs-Workflows |@@ -44,7 +45,11 @@ Der Skill enthält sofort nutzbare Vorlagen:- `feature-list.json` — JSON Schema und Beispiel-Feature-Liste- `init.sh` — Standard-Initialisierungsskript- `progress.md` — Vorlage für Session-Fortschrittslogs-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — Session-Handoff-Vorlage++### Scripts++Der Skill enthält außerdem reine Node.js-Scripts für Scaffold, Validierung, HTML-Assessment-Berichte und strukturelle Benchmark-Berichte.### Wie dieser Skill gebaut wurdediff --git a/docs/en/skills/index.md b/docs/en/skills/index.mdindex 0c29457..950ca60 100644--- a/docs/en/skills/index.md+++ b/docs/en/skills/index.md@@ -25,11 +25,12 @@ To use it with Claude Code, copy the `harness-creator/` directory into your proj### Reference Patterns-The skill includes 6 deep-dive reference documents:+The skill includes 7 focused reference documents:| Pattern | When to Use ||---------|-------------|| Memory Persistence | Agent forgets between sessions |+| Skill Runtime | Package reusable workflows as skills || Context Engineering | Context budget management, JIT loading || Tool Registry | Tool safety, concurrency control || Multi-Agent Coordination | Parallelism, specialization workflows |@@ -46,6 +47,10 @@ The skill bundles ready-to-use templates:- `progress.md` — Session progress log template- `session-handoff.md` — Session handoff template+### Scripts++The skill also includes plain Node.js scripts for scaffolding, validation, HTML assessment reports, and structural benchmark reports.+### How This Skill Was Built`harness-creator` was developed using the **skill-creator** methodology — Anthropic's official meta-skill for creating, testing, and iterating on agent skills. The skill-creator provides a structured workflow (draft → test → evaluate → iterate) with built-in eval runners, graders, and a benchmark viewer.diff --git a/docs/es/skills/index.md b/docs/es/skills/index.mdindex c1f896f..46571d9 100644--- a/docs/es/skills/index.md+++ b/docs/es/skills/index.md@@ -25,11 +25,12 @@ Para usarlo con Claude Code, copia el directorio `harness-creator/` a la ruta de### Patrones de referencia-El skill incluye 6 documentos de referencia detallados:+El skill incluye 7 documentos de referencia enfocados:| Patrón | Cuándo usarlo ||---------|-------------|| Memory Persistence | El agente olvida entre sesiones |+| Skill Runtime | Empaquetar workflows reutilizables como skills || Context Engineering | Gestión del presupuesto de contexto, carga JIT || Tool Registry | Seguridad de herramientas, control de concurrencia || Multi-Agent Coordination | Paralelismo, flujos de especialización |@@ -44,7 +45,11 @@ El skill incluye plantillas listas para usar:- `feature-list.json` — JSON Schema y ejemplo de lista de funciones- `init.sh` — script estándar de inicialización- `progress.md` — plantilla de registro de progreso de sesión-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — plantilla de handoff de sesión++### Scripts++El skill también incluye scripts de Node.js puro para scaffold, validación, informes HTML de evaluación y benchmarks estructurales.### Cómo se construyó este skilldiff --git a/docs/fr/skills/index.md b/docs/fr/skills/index.mdindex 9dd6fc5..c1dbd4b 100644--- a/docs/fr/skills/index.md+++ b/docs/fr/skills/index.md@@ -25,11 +25,12 @@ Pour l'utiliser avec Claude Code, copiez le dossier `harness-creator/` dans le c### Patterns de référence-Le skill inclut 6 documents de référence détaillés :+Le skill inclut 7 documents de référence ciblés :| Pattern | Quand l'utiliser ||---------|------------------|| Memory Persistence | L'agent oublie entre les sessions |+| Skill Runtime | Packager des workflows réutilisables en skills || Context Engineering | Gestion du budget de contexte, chargement JIT || Tool Registry | Sécurité des outils, contrôle de concurrence || Multi-Agent Coordination | Parallélisme, workflows de spécialisation |@@ -44,7 +45,11 @@ Le skill fournit des modèles prêts à l'emploi :- `feature-list.json` — JSON Schema et exemple de liste de fonctionnalités- `init.sh` — script d'initialisation standard- `progress.md` — modèle de journal de progression de session-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — modèle de handoff de session++### Scripts++Le skill inclut aussi des scripts Node.js purs pour le scaffold, la validation, les rapports HTML d'évaluation et les benchmarks structurels.### Comment ce skill a été construitdiff --git a/docs/ja/skills/index.md b/docs/ja/skills/index.mdindex cd2cc60..ad459eb 100644--- a/docs/ja/skills/index.md+++ b/docs/ja/skills/index.md@@ -25,11 +25,12 @@ Claude Code で使う場合は、`harness-creator/` ディレクトリをプロ### 参考パターン-このスキルには 6 つの詳細リファレンスがあります。+このスキルには 7 つの焦点化されたリファレンスがあります。| パターン | 使う場面 ||---------|----------|| Memory Persistence | エージェントがセッション間で忘れてしまう |+| Skill Runtime | 再利用可能なワークフローをスキルとしてまとめる || Context Engineering | コンテキスト予算管理、必要時ロード || Tool Registry | ツール安全性、並行制御 || Multi-Agent Coordination | 並列化、専門化ワークフロー |@@ -44,7 +45,11 @@ Claude Code で使う場合は、`harness-creator/` ディレクトリをプロ- `feature-list.json` — JSON Schema と機能リスト例- `init.sh` — 標準初期化スクリプト- `progress.md` — セッション進捗ログテンプレート-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — セッション引き継ぎテンプレート++### スクリプト++このスキルには、scaffold、検証、HTML 評価レポート、構造 benchmark レポート用の純粋な Node.js スクリプトも含まれています。### このスキルの作り方diff --git a/docs/ko/skills/index.md b/docs/ko/skills/index.mdindex ea5510f..78c6789 100644--- a/docs/ko/skills/index.md+++ b/docs/ko/skills/index.md@@ -25,11 +25,12 @@ Claude Code와 함께 사용하려면 `harness-creator/` 디렉터리를 프로### 참고 패턴 (Reference Patterns)-이 스킬에는 6가지 심층 참고 문서가 포함됩니다.+이 스킬에는 7개의 핵심 참고 문서가 포함되어 있습니다.| 패턴 (Pattern) | 사용 시점 (When to Use) ||---------|-------------|| Memory Persistence | 에이전트가 세션 간에 기억을 잃을 때 |+| Skill Runtime | 재사용 가능한 워크플로우를 스킬로 패키징 || Context Engineering | 컨텍스트 예산 관리, JIT 로딩 || Tool Registry | 도구 안전성, 동시성 제어 || Multi-Agent Coordination | 병렬화, 특수화 워크플로우 |@@ -44,7 +45,11 @@ Claude Code와 함께 사용하려면 `harness-creator/` 디렉터리를 프로- `feature-list.json` — JSON Schema와 기능 목록 예시- `init.sh` — 표준 초기화(initialization) 스크립트- `progress.md` — 세션 진행 로그 템플릿-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — 세션 핸드오프 템플릿++### 스크립트++이 스킬에는 scaffold, 검증, HTML 평가 리포트, 구조적 benchmark 리포트를 위한 순수 Node.js 스크립트도 포함되어 있습니다.### 이 스킬이 만들어진 방법 (How This Skill Was Built)diff --git a/docs/ru/skills/index.md b/docs/ru/skills/index.mdindex 41a422c..b0d3e2c 100644--- a/docs/ru/skills/index.md+++ b/docs/ru/skills/index.md@@ -25,11 +25,12 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator### Референсные паттерны-Скил включает 6 углублённых референсных документов:+Скил включает 7 сфокусированных референсных документов:| Паттерн | Когда применять ||---------|-------------|| Memory Persistence | Агент забывает между сессиями |+| Skill Runtime | Упаковка переиспользуемых workflow как skills || Context Engineering | Управление бюджетом контекста, JIT-загрузка || Tool Registry | Безопасность инструментов, контроль конкурентности || Multi-Agent Coordination | Параллелизм, флоу со специализацией |@@ -44,7 +45,11 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator- `feature-list.json` — JSON Schema + пример списка фич- `init.sh` — стандартный инициализационный скрипт- `progress.md` — шаблон лога прогресса сессии-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — шаблон handoff сессии++### Скрипты++Скил также включает чистые Node.js-скрипты для scaffold, валидации, HTML-отчётов оценки и структурных benchmark-отчётов.### Как был построен этот скилdiff --git a/docs/tr/skills/index.md b/docs/tr/skills/index.mdindex 6168782..f94ef96 100644--- a/docs/tr/skills/index.md+++ b/docs/tr/skills/index.md@@ -25,11 +25,12 @@ Claude Code ile kullanmak için `harness-creator/` dizinini projenizin yetenek y### Referans desenleri-Yetenek, 6 adet derinlemesine referans belge içerir:+Yetenek, 7 odaklı referans belge içerir:| Desen | Ne zaman kullanılır ||---------|-------------|| Bellek Kalıcılığı | Ajan oturumlar arasında unutuyor |+| Skill Runtime | Yeniden kullanılabilir iş akışlarını skill olarak paketleme || Bağlam Mühendisliği | Bağlam bütçesi yönetimi, JIT yükleme || Araç Kayıt Defteri | Araç güvenliği, eşzamanlılık kontrolü || Çoklu-Ajan Koordinasyonu | Paralellik, uzmanlaşma akışları |@@ -44,7 +45,11 @@ Yetenek, doğrudan kullanılabilen şablonlar içerir:- `feature-list.json` — JSON Schema + örnek özellik listesi- `init.sh` — Standart başlatma betiği- `progress.md` — Oturum ilerleme günlüğü şablonu-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — Oturum devir şablonu++### Scriptler++Yetenek ayrıca scaffold, doğrulama, HTML değerlendirme raporu ve yapısal benchmark raporu için saf Node.js scriptleri içerir.### Bu yetenek nasıl geliştirildidiff --git a/docs/uz/skills/index.md b/docs/uz/skills/index.mdindex 0f4ed20..1c3db15 100644--- a/docs/uz/skills/index.md+++ b/docs/uz/skills/index.md@@ -25,11 +25,12 @@ Undan Claude Code bilan foydalanish uchun, `harness-creator/` jildini loyihangiz### Maʼlumotnoma andozalari (Reference Patterns)-Ushbu koʻnikma oʻz ichiga chuqur tahlil qilingan 6 ta maʼlumotnoma hujjatini (reference documents) oladi:+Ushbu koʻnikma oʻz ichiga 7 ta yoʻnaltirilgan maʼlumotnoma hujjatini (reference documents) oladi:| Andoza (Pattern) | Qachon foydalanish kerak ||---------|-------------|| Xotirani saqlab qolish (Memory Persistence) | Agent sessiyalar oʻrtasida unutib qoʻysa |+| Skill Runtime | Qayta ishlatiladigan workflowʼlarni skill sifatida paketlash || Kontekst muhandisligi (Context Engineering) | Kontekst byudjetini boshqarish, kerak vaqtda (JIT) yuklash || Vositalar reyestri (Tool Registry) | Vositalar xavfsizligi, parallellik nazorati (concurrency control) || Koʻp agentli muvofiqlashtirish (Multi-Agent Coordination) | Parallellik, ixtisoslashish (specialization) jarayonlari |@@ -44,7 +45,11 @@ Koʻnikma oʻz ichiga foydalanishga tayyor andozalarni oladi:- `feature-list.json` — JSON Schema + namunaviy funksiyalar roʻyxati (example feature list)- `init.sh` — Standart inisializatsiya skripti- `progress.md` — Sessiya jarayoni jurnali (progress log) andozasi-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — Sessiyani topshirish andozasi++### Skriptlar++Koʻnikma scaffold, tekshirish, HTML baholash hisoboti va strukturaviy benchmark hisoboti uchun sof Node.js skriptlarini ham oʻz ichiga oladi.### Ushbu koʻnikma qanday yaratilgan (How This Skill Was Built)diff --git a/docs/vi/skills/index.md b/docs/vi/skills/index.mdindex 25d08a7..5e61b63 100644--- a/docs/vi/skills/index.md+++ b/docs/vi/skills/index.md@@ -25,11 +25,12 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator### Các Mẫu Tham Khảo (Reference Patterns)-Kỹ năng này bao gồm 6 tài liệu tham khảo chuyên sâu:+Kỹ năng này bao gồm 7 tài liệu tham khảo tập trung:| Mẫu (Pattern) | Khi nào sử dụng ||---------|-------------|| Lưu trữ bộ nhớ | Agent quên giữa các phiên |+| Skill Runtime | Đóng gói workflow tái sử dụng thành skill || Kỹ thuật ngữ cảnh | Quản lý ngân sách ngữ cảnh, nạp tức thời (JIT) || Đăng ký công cụ | An toàn công cụ, kiểm soát đồng thời || Điều phối Đa Agent | Xử lý song song, quy trình làm việc chuyên môn hóa |@@ -44,7 +45,11 @@ Kỹ năng này đi kèm với các mẫu sẵn sàng sử dụng:- `feature-list.json` — JSON Schema + danh sách tính năng mẫu- `init.sh` — Kịch bản khởi tạo tiêu chuẩn- `progress.md` — Mẫu nhật ký tiến độ phiên làm việc-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — Mẫu bàn giao phiên++### Scripts++Kỹ năng này cũng bao gồm các script Node.js thuần cho scaffold, xác thực, báo cáo đánh giá HTML và báo cáo benchmark cấu trúc.### Kỹ năng này được Xây dựng như thế nàodiff --git a/docs/zh-TW/skills/index.md b/docs/zh-TW/skills/index.mdindex f97d7fa..907b576 100644--- a/docs/zh-TW/skills/index.md+++ b/docs/zh-TW/skills/index.md@@ -25,11 +25,12 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator### 參考模式-技能包含 6 個深入的模式參考檔案:+技能包含 7 個聚焦的模式參考檔案:| 模式 | 適用場景 ||------|----------|| 記憶持久化 | agent 在會話間遺忘專案知識 |+| Skill Runtime | 將可重用 workflow 打包成技能 || 上下文工程 | 上下文預算管理、按需載入、委託隔離 || 工具註冊 | 工具安全、併發控制、權限管道 || 多智慧體協調 | 平行化、專業化、研究員→實作者工作流 |@@ -44,7 +45,11 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator- `feature-list.json` — JSON Schema + 功能列表示例- `init.sh` — 標準初始化腳本- `progress.md` — 會話進度日誌範本-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — 會話交接範本++### 腳本++技能也包含純 Node.js 腳本,可用於腳手架生成、校驗、HTML 評估報告與結構化 benchmark 報告。### 開發過程diff --git a/docs/zh/skills/index.md b/docs/zh/skills/index.mdindex 166a522..6850629 100644--- a/docs/zh/skills/index.md+++ b/docs/zh/skills/index.md@@ -25,11 +25,12 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator### 参考模式-技能包含 6 个深入的模式参考文档:+技能包含 7 个聚焦的模式参考文档:| 模式 | 适用场景 ||------|----------|| 记忆持久化 | agent 在会话间遗忘项目知识 |+| Skill Runtime | 将可复用 workflow 打包成技能 || 上下文工程 | 上下文预算管理、按需加载、委托隔离 || 工具注册 | 工具安全、并发控制、权限管道 || 多智能体协调 | 并行化、专业化、研究员→实施者工作流 |@@ -44,7 +45,11 @@ npx skills add walkinglabs/learn-harness-engineering --skill harness-creator- `feature-list.json` — JSON Schema + 功能列表示例- `init.sh` — 标准初始化脚本- `progress.md` — 会话进度日志模板-- `session-handoff.md` — Session handoff template+- `session-handoff.md` — 会话交接模板++### 脚本++技能也包含纯 Node.js 脚本,可用于脚手架生成、校验、HTML 评估报告与结构化 benchmark 报告。### 开发过程diff --git a/skills/README-AR.md b/skills/README-AR.mdindex 6afdc13..9e4a2d2 100644--- a/skills/README-AR.md+++ b/skills/README-AR.md@@ -21,10 +21,10 @@Skill لإنتاج harness engineering بجودة إنتاجية لوكلاء البرمجة. تساعد على إنشاء وتقييم وتحسين ملفات harness الخاصة بالوكيل، مثل AGENTS.md وقوائم الميزات وسير عمل التحقق وآليات استمرار الجلسة.-- **5 أنماط مرجعية**: Memory Persistence وContext Engineering وTool Registry وMulti-Agent Coordination وLifecycle & Bootstrap-- **قوالب**: AGENTS.md وfeature-list.json وinit.sh وprogress.md-- **5 حالات eval مدمجة**-- **ثنائية اللغة**: English + 中文+- **7 أنماط مرجعية**: Memory Persistence وSkill Runtime وContext Engineering وTool Registry وMulti-Agent Coordination وLifecycle & Bootstrap وGotchas+- **قوالب**: AGENTS.md/CLAUDE.md وfeature-list.json وinit.sh وprogress.md وsession-handoff.md+- **Scripts**: إنشاء scaffold، التحقق، تقرير HTML، وbenchmark بنيوي+- **10 حالات eval مدمجة**راجع التوثيق الكامل في [harness-creator/README.md](harness-creator/README.md).@@ -49,7 +49,8 @@ Skill لإنتاج harness engineering بجودة إنتاجية لوكلاء ا- لا توجد backdoors أو URL مخفية أو payloads مشفرة- لا يوجد تسريب بيانات أو بيانات اعتماد hardcoded- لا توجد ثغرات command injection-- `init.sh` يشغل فقط أوامر npm lifecycle القياسية+- تستخدم scripts وحدات Node.js المدمجة فقط+- `init.sh` المولّد يشغل أوامر التحقق المكتشفة للمشروع## الرخصةdiff --git a/skills/README-DE.md b/skills/README-DE.mdindex ea488f9..c1e9d39 100644--- a/skills/README-DE.md+++ b/skills/README-DE.md@@ -21,10 +21,10 @@ Dieses Verzeichnis enthält wiederverwendbare AI agent skills für das Projekt LProduction-orientierter Harness-Engineering-Skill für AI coding agents. Er hilft beim Erstellen, Bewerten und Verbessern von Agent-Harness-Dateien wie AGENTS.md, Feature-Listen, Verification Workflows und Mechanismen für Session Continuity.-- **5 Referenzmuster**: Memory Persistence, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap-- **Templates**: AGENTS.md, feature-list.json, init.sh, progress.md-- **5 integrierte Eval-Testfälle**-- **Zweisprachig**: English + 中文+- **7 Referenzmuster**: Memory Persistence, Skill Runtime, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap, Gotchas+- **Templates**: AGENTS.md/CLAUDE.md, feature-list.json, init.sh, progress.md, session-handoff.md+- **Scripts**: Scaffold, Validierung, HTML-Assessment und struktureller Benchmark+- **10 integrierte Eval-Testfälle**Siehe [harness-creator/README.md](harness-creator/README.md) für die vollständige Dokumentation.@@ -49,7 +49,8 @@ Alle Dateien in diesem Verzeichnis wurden geprüft:- Keine Backdoors, versteckten URLs oder codierten Payloads- Keine Datenexfiltration oder hardcodierten Zugangsdaten- Keine Command-Injection-Schwachstellen-- `init.sh` führt nur Standardbefehle aus dem npm lifecycle aus+- Scripts verwenden nur eingebaute Node.js-Module+- Generiertes `init.sh` führt erkannte Projekt-Verifikationsbefehle aus## Lizenzdiff --git a/skills/README-ES.md b/skills/README-ES.mdindex 2adde3c..163bc45 100644--- a/skills/README-ES.md+++ b/skills/README-ES.md@@ -21,10 +21,10 @@ Este directorio contiene skills reutilizables para agentes de IA en el proyectoSkill de harness engineering de nivel producción para agentes de programación. Ayuda a crear, evaluar y mejorar archivos de harness para agentes, como AGENTS.md, listas de funciones, flujos de verificación y mecanismos de continuidad entre sesiones.-- **5 patrones de referencia**: Memory Persistence, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap-- **Plantillas**: AGENTS.md, feature-list.json, init.sh, progress.md-- **5 casos de eval integrados**-- **Bilingüe**: English + 中文+- **7 patrones de referencia**: Memory Persistence, Skill Runtime, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap, Gotchas+- **Plantillas**: AGENTS.md/CLAUDE.md, feature-list.json, init.sh, progress.md, session-handoff.md+- **Scripts**: scaffold, validación, informe HTML y benchmark estructural+- **10 casos de eval integrados**Consulta la documentación completa en [harness-creator/README.md](harness-creator/README.md).@@ -49,7 +49,8 @@ Los archivos de este directorio fueron revisados:- Sin backdoors, URL ocultas ni payloads codificados- Sin exfiltración de datos ni credenciales hardcodeadas- Sin vulnerabilidades de command injection-- `init.sh` solo ejecuta comandos estándar del ciclo de vida de npm+- Los scripts usan solo módulos integrados de Node.js+- El `init.sh` generado ejecuta los comandos de verificación detectados del proyecto## Licenciadiff --git a/skills/README-FR.md b/skills/README-FR.mdindex 32d86ad..6b51c29 100644--- a/skills/README-FR.md+++ b/skills/README-FR.md@@ -21,10 +21,10 @@ Ce dossier contient des skills réutilisables pour agents IA dans le projet LearSkill de harness engineering de niveau production pour agents de codage IA. Elle aide à créer, évaluer et améliorer les fichiers de harness d'agent, notamment AGENTS.md, les listes de fonctionnalités, les workflows de vérification et les mécanismes de continuité de session.-- **5 modèles de référence** : Memory Persistence, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap-- **Templates** : AGENTS.md, feature-list.json, init.sh, progress.md-- **5 cas de test eval intégrés**-- **Bilingue** : English + 中文+- **7 modèles de référence** : Memory Persistence, Skill Runtime, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap, Gotchas+- **Templates** : AGENTS.md/CLAUDE.md, feature-list.json, init.sh, progress.md, session-handoff.md+- **Scripts** : scaffold, validation, rapport HTML et benchmark structurel+- **10 cas de test eval intégrés**Voir la documentation complète : [harness-creator/README.md](harness-creator/README.md).@@ -49,7 +49,8 @@ Les fichiers de ce dossier ont été audités :- Aucun backdoor, URL cachée ou payload encodé- Aucune exfiltration de données ni identifiants codés en dur- Aucune vulnérabilité de command injection-- `init.sh` exécute uniquement des commandes npm lifecycle standard+- Les scripts utilisent uniquement les modules intégrés de Node.js+- Le `init.sh` généré exécute les commandes de vérification détectées du projet## Licencediff --git a/skills/README-JA.md b/skills/README-JA.mdindex 6e4d4ae..427ffea 100644--- a/skills/README-JA.md+++ b/skills/README-JA.md@@ -21,10 +21,10 @@AI コーディング agent 向けの本番品質 harness engineering skill です。AGENTS.md、機能リスト、検証ワークフロー、セッション継続メカニズムなどの agent harness ファイルの作成、評価、改善を支援します。-- **5 つの参照パターン**:Memory Persistence、Context Engineering、Tool Registry、Multi-Agent Coordination、Lifecycle & Bootstrap-- **テンプレート**:AGENTS.md、feature-list.json、init.sh、progress.md-- **5 つの組み込み eval テストケース**-- **バイリンガル**:English + 中文+- **7 つの参照パターン**:Memory Persistence、Skill Runtime、Context Engineering、Tool Registry、Multi-Agent Coordination、Lifecycle & Bootstrap、Gotchas+- **テンプレート**:AGENTS.md/CLAUDE.md、feature-list.json、init.sh、progress.md、session-handoff.md+- **スクリプト**:scaffold、検証、HTML 評価レポート、構造 benchmark+- **10 個の組み込み eval テストケース**詳細は [harness-creator/README.md](harness-creator/README.md) を参照してください。@@ -49,7 +49,8 @@ Skills は progressive disclosure を使います。agent は最初に名前と- バックドア、隠し URL、エンコードされた payload はありません- データ流出やハードコードされた認証情報はありません- コマンドインジェクション脆弱性はありません-- `init.sh` は標準 npm lifecycle コマンドのみを実行します+- スクリプトは Node.js の組み込みモジュールのみを使用します+- 生成された `init.sh` は検出されたプロジェクト検証コマンドを実行します## ライセンスdiff --git a/skills/README-KO.md b/skills/README-KO.mdindex 46b3f54..b193efd 100644--- a/skills/README-KO.md+++ b/skills/README-KO.md@@ -23,10 +23,10 @@AI 코딩 에이전트를 위한 프로덕션 하네스 엔지니어링 스킬입니다. 에이전트 하네스 파일(AGENTS.md, 기능 목록, 검증 워크플로우, 세션 연속성(session continuity) 메커니즘)을 생성하고 평가하며 개선하는 데 도움을 줍니다.-- **5가지 참고 패턴**: 메모리 영속성(Memory Persistence), 컨텍스트 엔지니어링(Context Engineering), 도구 레지스트리(Tool Registry), 다중 에이전트 조율(Multi-Agent Coordination), 라이프사이클 및 부트스트랩(Lifecycle & Bootstrap)-- **템플릿**: AGENTS.md, feature-list.json, init.sh, progress.md-- **5개의 내장 평가(eval) 테스트 케이스**-- **이중 언어**: 영어 + 中文+- **7가지 참고 패턴**: 메모리 영속성(Memory Persistence), 스킬 런타임(Skill Runtime), 컨텍스트 엔지니어링(Context Engineering), 도구 레지스트리(Tool Registry), 다중 에이전트 조율(Multi-Agent Coordination), 라이프사이클 및 부트스트랩(Lifecycle & Bootstrap), Gotchas+- **템플릿**: AGENTS.md/CLAUDE.md, feature-list.json, init.sh, progress.md, session-handoff.md+- **스크립트**: scaffold, 검증, HTML 평가 리포트, 구조적 benchmark+- **10개의 내장 평가(eval) 테스트 케이스**전체 문서는 [harness-creator/README.md](harness-creator/README.md)를 참조하십시오.@@ -54,10 +54,12 @@ skills/├── README-VI.md # 베트남어 버전├── README-DE.md # 독일어 버전└── harness-creator/ # 하네스 엔지니어링 스킬- ├── SKILL.md # 주요 스킬 정의 (이중 언어)+ ├── SKILL.md # 주요 스킬 정의├── SKILL.md.en # 영어 전용 버전├── README.md # 상세 문서├── metadata.json # 스킬 메타데이터 및 트리거+ ├── agents/ # 스킬 UI 메타데이터+ ├── scripts/ # 스캐폴드, 검증, benchmark 보조 스크립트├── evals/ # 테스트 케이스├── templates/ # 스캐폴드(scaffold) 템플릿└── references/ # 심층 패턴 문서@@ -82,7 +84,8 @@ skills/- 백도어, 숨겨진 URL, 인코딩된 페이로드 없음- 데이터 유출 또는 하드코딩된 자격 증명 없음- 명령 주입 취약점 없음-- `init.sh`는 표준 npm 라이프사이클 명령만 실행+- 스크립트는 Node.js 내장 모듈만 사용+- 생성된 `init.sh`는 감지된 프로젝트 검증 명령을 실행## 라이선스diff --git a/skills/README-VI.md b/skills/README-VI.mdindex fad7fe9..ff7e2d2 100644--- a/skills/README-VI.md+++ b/skills/README-VI.md@@ -21,10 +21,10 @@ Thu muc nay chua cac AI agent skills co the tai su dung cho du an Learn HarnessSkill harness engineering cap production cho AI coding agents. Skill nay giup tao, danh gia va cai thien cac file harness cua agent, gom AGENTS.md, danh sach tinh nang, verification workflows va co che tiep noi phien lam viec.-- **5 mau tham chieu**: Memory Persistence, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap-- **Templates**: AGENTS.md, feature-list.json, init.sh, progress.md-- **5 eval test case tich hop**-- **Song ngu**: English + 中文+- **7 mau tham chieu**: Memory Persistence, Skill Runtime, Context Engineering, Tool Registry, Multi-Agent Coordination, Lifecycle & Bootstrap, Gotchas+- **Templates**: AGENTS.md/CLAUDE.md, feature-list.json, init.sh, progress.md, session-handoff.md+- **Scripts**: scaffold, validate, bao cao HTML, structural benchmark+- **10 eval test case tich hop**Xem tai lieu day du tai [harness-creator/README.md](harness-creator/README.md).@@ -49,7 +49,8 @@ Tat ca file trong thu muc nay da duoc kiem tra bao mat:- Khong co backdoor, URL an hoac payload duoc ma hoa- Khong ro ri du lieu hoac credential hardcoded- Khong co lo hong command injection-- `init.sh` chi chay cac lenh npm lifecycle tieu chuan+- Scripts chi dung module co san cua Node.js+- `init.sh` duoc tao se chay cac lenh verification phat hien tu project## Giay phepdiff --git a/skills/README-ZH-TW.md b/skills/README-ZH-TW.mdindex a0c069a..217e876 100644--- a/skills/README-ZH-TW.md+++ b/skills/README-ZH-TW.md@@ -21,10 +21,10 @@面向 AI 編程 agent 的生產級 harness engineering 技能。它協助建立、評估與改進 agent harness 檔案,例如 AGENTS.md、功能清單、驗證工作流與跨會話延續機制。-- **5 個參考模式**:記憶持久化、上下文工程、工具註冊、多 agent 協調、生命週期與啟動-- **範本**:AGENTS.md、feature-list.json、init.sh、progress.md-- **5 個內建 eval 測試案例**-- **雙語支援**:English + 中文+- **7 個參考模式**:記憶持久化、技能執行期、上下文工程、工具註冊、多 agent 協調、生命週期與啟動、常見陷阱+- **範本**:AGENTS.md/CLAUDE.md、feature-list.json、init.sh、progress.md、session-handoff.md+- **腳本**:腳手架生成、校驗、HTML 評估報告、結構化 benchmark+- **10 個內建 eval 測試案例**完整文件請見 [harness-creator/README.md](harness-creator/README.md)。@@ -49,7 +49,8 @@- 無後門、隱藏 URL 或編碼 payload- 無資料外洩或硬編碼憑證- 無命令注入漏洞-- `init.sh` 僅執行標準 npm lifecycle 命令+- 腳本僅使用 Node.js 內建模組+- 生成的 `init.sh` 會執行偵測到的專案驗證命令## 授權