HoangNguyen0403/agent-skills-standard

What it solves

AI coding agents often struggle with team-specific coding standards, leading to generic code or the need for repetitive, token-heavy prompting. This project provides a portable, version-controlled layer of engineering rules that AI agents can load on demand, reducing token waste and ensuring consistency across different AI tools.

How it works

It uses a hierarchical loading system to avoid "prompt loss" and save tokens. Instead of injecting all rules into every prompt, it follows a resolution path: Router Table (AGENTS.md) $\rightarrow$ Category Index (_INDEX.md) $\rightarrow$ Specific Skill (SKILL.md). The AI only loads the rules that match the file being edited or the task at hand.

The system is managed via a CLI (ags) for distribution and an optional MCP server for runtime enforcement, ensuring agents actually call the necessary skills during a session.

Who it’s for

Software engineering teams who use AI coding agents (such as Cursor, Claude Code, GitHub Copilot, Gemini, Windsurf, and others) and want to synchronize their SDLC workflows and coding standards across their entire team.

Highlights

• Multi-Agent Support: Generates native assets for a wide variety of AI tools including Cursor, Claude, Copilot, and Gemini.
• Token Efficiency: Optimized skills average ~500 tokens, using up to 85% fewer tokens than traditional prompt engineering.
• Extensive Library: Includes nearly 300 ready-to-use skills across 20+ frameworks (e.g., Flutter, React, Go, Python, NestJS).
• SDLC Workflow Spine: Provides a native lifecycle for agents to follow, covering stages from brainstorming and planning to security testing and verification.
• Zero-Trust Architecture: Implements a hierarchical resolution system to prevent AI confusion and prompt loss.