Integrate with Skills

Use a coding agent that supports Agent Skills to accelerate requirement analysis, integration, and cross-platform troubleshooting for AI Chat. With Tange Skills configured, the agent can use the current official documentation and your existing projects to select an appropriate implementation and validation workflow.

$ai-chat coordinates role configuration, the application server, device sessions, and events, and uses $tirtc when connection or media work is required. A Skill is not an SDK and does not run as a separate program.

Install Tange Skills

Before you begin, prepare Node.js 22.20.0 or later, Python 3.10 or later, and curl. Also make sure:

• Your development tool supports Agent Skills; the example below uses Codex.
• Your network can access the official AI Chat and TiRTC documentation, SDKs, and example downloads.

Configure the Skills with the Skills CLI:

npx --yes skills@latest add \
  https://github.com/tangeai/skills.git \
  --skill tirtc \
  --skill ai-chat \
  -g -a codex -y

-g installs the Skills for the current system user; remove it to install only in the current project. -a codex selects Codex; replace it with another agent supported by the Skills CLI when needed.

When the command finishes, start a new task or refresh the Skill list.

Repository

• Tange Skills

See the repository README for the latest usage and update instructions.

Best Practices

After configuration, ask the coding agent to integrate AI Chat or troubleshoot a cross-platform issue. To obtain actionable and verifiable results:

• Specify AI Chat, the involved endpoints, target conversation flow, and required behavior. Name only the primary Skill, $ai-chat; it coordinates $tirtc when needed.
• Provide the role configuration, each project location, current implementation, allowed changes, real-device conditions, and completion criteria.
• For troubleshooting, include the observed behavior, reproduction steps, error codes, and logs from the application server and device.
• Do not put AccessKeyId, SecretKeyId, device secrets, tokens, or third-party agent keys in the prompt. Ask the agent how to provide credentials through secure configuration.
• Ask the agent to run existing builds and tests and verify captions or response audio, interruption or device events, cleanup after session end, and a subsequent session. Clearly mark checks that require a real device, valid credentials, or agent configuration as pending.

For manual steps, continue with Overall Flow, AI Talk Console, Server Integration, Device Integration, and Event Protocol.

Prompt Example

Send the following prompt to the coding agent:

Use $ai-chat to add AI Chat to the existing TiRTC device project.
First check the current official documentation, role configuration, and existing
project. Follow the current architecture to implement the server-side token,
device session, captions, and voice loop. Run the existing build and tests,
then use a real device to verify that a conversation can start, be interrupted,
and end.