翻訳待ち：Show HN: TechSkills – Open-source skill modules for AI coding agents

AI サービスが一時的に利用できないため、復旧後に翻訳を補完します。

Notifications You must be signed in to change notification settings Fork 0 Star 4 BranchesTags Open more actions menu Latest commit History 3 Commits 3 Commits Folders and files NameName Last commit message Last commit date skills skills .gitignore .gitignore CONTRIBUTING.md CONTRIBUTING.md LICENSE.txt LICENSE.txt README.md README.md Repository files navigation A curated, open-source library of skills for AI coding agents. AI agents are powerful generalists, but they often lack the deep procedural knowledge, opinionated best practices, and domain-specific patterns that experienced software engineers carry. TechSkills-Lib bridges that gap — it provides ready-to-use skill modules covering frontend engineering, backend development, system design, and more, so any AI agent can deliver expert-level output across the full software development stack. Agent-agnostic, framework-neutral, and community-driven. What Are Skills? Skills are simply SKILL.md files (with optional supporting references and assets) that give an AI agent context, knowledge, and step-by-step workflows for a specific domain. When an agent loads a skill, it gains structured instructions, design checklists, code patterns, and quality standards — the kind of expertise that turns a generic AI response into production-quality engineering work. Any AI coding agent that supports loading markdown-based instructions can use these skills out of the box. No framework lock-in. No vendor lock-in. Any AI agent can use these skills. Why? AI agents are smart generalists, but they lack the procedural knowledge and opinionated best practices that experienced engineers carry. This library fills that gap with: Battle-tested patterns — not theory, but production-ready approaches Progressive disclosure — lean core instructions, detailed references loaded only when needed Framework-neutral design — pseudocode patterns that adapt to any stack Quality gates — eval cases to verify skill effectiveness Available Skills Skill Description frontend-engineer High-end UI implementation — responsive layouts, accessibility, visual polish, design systems, React/Vue/Svelte/React Native backend-engineer Production-grade APIs, auth, databases, queues, reliability, observability, testing, security Skill Structure Every skill follows the same structure: skill-name/ ├── SKILL.md # Core instructions (required) ├── evals/ │ └── evals.json # Eval cases for quality verification ├── references/ # Detailed guides, loaded on-demand │ ├── checklist-a.md │ └── checklist-b.md └── assets/ # Starter templates, patterns, snippets └── starter/ How Skills Work Skills use a three-level loading system to manage context efficiently: Metadata (name + description in YAML frontmatter) — always visible to the agent (~100 words). This is what triggers the skill. SKILL.md body — loaded when the skill activates. Kept lean (under 500 lines). References & Assets — loaded only when the agent determines they're needed. Unlimited depth. This means skills don't waste context window space. The agent loads only what it needs, when it needs it. How to Use Option 1: Copy a skill into your agent's skill directory # Clone the repo git clone https://github.com/your-username/techskills-lib.git # Copy the skill you need cp -r techskills-lib/skills/frontend-engineer ~/.your-agent/skills/ Option 2: Reference directly Point your AI agent's skill path to this repository's skills/ directory. Option 3: Cherry-pick references Don't need the full skill? Copy individual reference files (e.g., database-patterns.md, accessibility-checklist.md) into your project docs. Contributing We welcome contributions! Whether it's a new skill, improvements to existing ones, or bug fixes in eval cases. See CONTRIBUTING.md for guidelines on how to add or improve skills. Quick Start for Contributors Fork the repo Create a skill folder under skills/ Add SKILL.md with YAML frontmatter (name + description) Add evals/evals.json with test cases Add references/ for detailed checklists (optional) Add assets/ for starter patterns (optional) Submit a PR Design Principles Agent-agnostic — no skill references a specific AI product. Works with any agent. Concise over verbose — every line must justify its token cost. Patterns over prescriptions — pseudocode and guidelines that adapt to any stack. Progressive disclosure — lean SKILL.md, heavy detail in references. Eval-driven quality — every skill includes eval cases to verify it works. License MIT © Debabrata Saha MIT license Contributing Contributing Activity Stars 4 stars Watchers 0 watching Forks 0 forks Report repository