shengwang-skills

Shengwang Skills

Reusable skills for AI coding agents building with the Shengwang platform. These skills help agents accurately integrate, configure, and debug Shengwang products.

Available Skills

Skill	Product	Description
conversational-ai	ConvoAI	AI voice agent lifecycle: create/stop/update/query. Supports Go, Java, Python
rtc	RTC SDK	Real-time audio/video calls. Web, Android, iOS, Flutter, and more
rtm	RTM	Real-time messaging, signaling, presence
cloud-recording	Cloud Recording	Server-side recording of RTC sessions
token-server	Token Server	Server-side token generation (AccessToken2)
general	General	Credential management, REST auth patterns

Quick Start

Installation Methods

Skills CLI

Install with CLI:

npx skills add Shengwang-Community/skills

This is the most direct installation method. After installation, restart the session or refresh the skills list according to your coding agent's instructions.

Claude Code Plugin Marketplace

Run the following command in Claude Code:

plugin marketplace add Shengwang-Community/skills

OpenClaw

Install via ClawHub. Use install for the initial installation and update for subsequent updates.

clawhub install voice-ai-integration
clawhub update voice-ai-integration

2. Download Doc Index (Recommended)

Download the documentation index for fetching latest API docs during development:

bash skills/voice-ai-integration/scripts/fetch-docs.sh

This saves the doc index to skills/voice-ai-integration/references/docs.txt. Skills use it to look up and fetch documentation directly via HTTP — no external server process needed.

Skills work without the doc index too — they fall back to local reference docs and external doc links.

3. Start Using

Describe your needs to the agent — skills trigger automatically:

• "I want to build an AI voice assistant" → ConvoAI + RTC integration
• "Generate an RTC token in Go" → Token Server module
• "How to implement video calls on Web" → RTC SDK module
• "Download the ConvoAI Go SDK" → Resource Downloader

How It Works

User Request
   │
   ▼
skills/voice-ai-integration/SKILL.md (entry point)
   │
   ├─ Clear request → Route directly to product module
   │
   └─ Vague request → Ask one clarifying question, then route

The entry point (skills/voice-ai-integration/SKILL.md) matches the request to a product module:

• Clear and actionable → route directly to the matching product module
• Vague or multi-product → ask one clarifying question, then route

Each product module follows a consistent workflow: confirm credentials → fetch latest docs → generate code → validate.

Repository Structure

shengwang-skills/
├── README.md                  # This file
├── AGENTS.md                  # Agent entry point instructions
├── CLAUDE.md                  # → AGENTS.md
├── CONTRIBUTING.md            # Contribution guidelines
├── scripts/
│   └── validate-skills.sh     # Link and frontmatter validation
├── tests/
│   └── eval-cases.md          # Evaluation test cases
└── skills/
    └── voice-ai-integration/     # The skill (agentskills.io standard)
        ├── SKILL.md               # Entry point and router (only SKILL.md)
        └── references/            # All product modules and shared knowledge
            ├── doc-fetching.md        # Doc fetching guide
            ├── docs.txt               # Local doc index
            ├── general/               # Credentials, REST auth
            ├── conversational-ai/     # ConvoAI
            ├── rtc/                   # RTC SDK
            ├── rtm/                   # RTM
            ├── cloud-recording/       # Cloud Recording
            └── token-server/          # Token generation

Design Philosophy

• Behavior over knowledge: skills teach agents how to approach integration; doc fetching provides specific APIs
• Single responsibility: each module does one thing
• Progressive disclosure: SKILL.md serves as navigation; detailed content lives in references/ and module README.md files
• Explicit failure paths: every module defines error handling
• Eval-driven iteration: validate changes against tests/eval-cases.md

Validation

bash scripts/validate-skills.sh

Checks all SKILL.md frontmatter format and markdown link validity.

Contributing

See CONTRIBUTING.md. Key requirements: