Agent Skills Repository

A curated collection of production-grade, modular AI Agent Skills designed to enhance AI coding assistants (such as Google Antigravity, Claude Code, and other agentic IDE environments) with specialized domain knowledge, best practices, checklists, and automated workflows.

📦 Available Skills

Skill	Version	Description	Target Formats / Tools
`c4-diagram`	`1.0.1`	Comprehensive software architecture visualization based on Simon Brown's Visualising Software Architecture. Generates Context, Container, Component, Deployment, Dynamic, and Landscape diagrams with Draw.io and Excalidraw MCP server integration.	Mermaid C4, PlantUML C4, Structurizr DSL, Draw.io, Excalidraw
`web-demo-generation`	`1.0.0`	Ingest natural language prompts for web application demos, plan markdown walkthrough scripts, and deterministically record high-definition demo videos with zero-LLM TTS voice narration, smooth cursor paths, click ripples, and spotlight highlights.	Playwright, edge-tts, pyttsx3, FFmpeg, Markdown DSL

🛠️ Skill Directory Architecture

Every skill in this repository follows a standardized, progressive-disclosure package layout:

skills/
└── <skill-name>/
    ├── SKILL.md                         # Main skill entrypoint (YAML frontmatter + instructions + triggers)
    ├── README.md                        # Skill-specific documentation and quickstart
    ├── LICENSE.md                       # License file
    ├── references/                      # Deep-dive documentation, vocabulary, rules, and syntax guides
    ├── examples/                        # Canonical, worked reference implementations & DSL models
    └── scripts/                         # Helper CLI scripts, linters, and verification tools

🚀 How Skills Work with AI Agents

1. Progressive Disclosure

Agent skills are designed to minimize context token overhead:

Only the skill's name and description from the YAML frontmatter in SKILL.md are indexed initially.
When an agent recognizes a relevant user request (e.g., "Create a C4 container diagram for my microservices"), it loads SKILL.md on-demand and traverses linked documents in references/ as needed.

2. Integration with Model Context Protocol (MCP)

Skills can orchestrate external MCP servers to produce real workspace artifacts. For example:

The c4-diagram skill generates clean Mermaid C4 or PlantUML C4 syntax and invokes the Draw.io MCP Server (drawio) to render editable .drawio and .svg visual diagram files in the project.
The portable Agent Plugin also configures the official Excalidraw MCP Server (excalidraw) for interactive hand-drawn diagrams.

3. Portable Agent Plugin releases

The repository-root plugin.json and mcp.json files define the Agent Plugins Specification v1.0.0 manifest and MCP configuration. The release workflow assembles these files with all bundled skills using the standard portable layout:

agent-skills-plugin/
├── plugin.json
├── mcp.json
├── LICENSE
└── skills/
    ├── c4-diagram/
    │   └── SKILL.md
    └── web-demo-generation/
        └── SKILL.md

Every pushed commit produces a SHA-tagged GitHub Release containing .tar.gz and .zip archives and checksums after validating plugin.json and mcp.json with apv.

🧪 Quality & Validation

Skills in this repository include automated validation tools to maintain high standards of quality and notation hygiene.

For example, to lint any C4 architecture diagram against Simon Brown's Appendix A Checklist:

cd c4-diagram
python scripts/validate_c4.py examples/ references/

📄 License

All skills in this repository are licensed under the MIT License.