Agent Skill

This repository ships a reusable coding-agent skill for implementing LIVE Studio Local Data Access in any application stack.

Install it in another project

The fastest distribution method is the open Skills CLI. It discovers the skill from this GitHub repository and installs it for the coding agents detected on the developer machine:

npx skills add liangpeiran-bit/tiktok-live-studio-local-data-access \
  --skill tiktok-live-studio-local-data-access \
  -g

-g makes the skill available across the developer's projects. Omit -g to install it only in the current project. To target particular agents non-interactively:

npx skills add liangpeiran-bit/tiktok-live-studio-local-data-access \
  --skill tiktok-live-studio-local-data-access \
  -g -a codex -a claude-code -a cursor -y

Update an installed copy after this repository changes:

npx skills update tiktok-live-studio-local-data-access -g -y

Developers without Node.js/npm can download the repository or a release archive and copy the entire directory into the target project:

source:
  .agents/skills/tiktok-live-studio-local-data-access/

recommended target for Codex, Cursor, OpenCode, Copilot, and Gemini CLI:
  <project>/.agents/skills/tiktok-live-studio-local-data-access/

Claude Code target:
  <project>/.claude/skills/tiktok-live-studio-local-data-access/

Trae target:
  <project>/.trae/skills/tiktok-live-studio-local-data-access/

Do not copy only SKILL.md; the protocol, events, stack adaptation, and verification references are part of the workflow.

For teams using several agent products, keep .agents/skills/ as the canonical copy. Prefer an explicit read prompt or a locally managed link/copy for Claude Code and Trae instead of committing multiple editable copies that can drift.

Give your agent its first task

Use the LIVE Studio Local Data Access skill to integrate live.like into this project. Keep the existing stack, start with mock tests, and do not request or print real credentials.

LLM READY

Documentation for AI coding assistants

Give these URLs to Codex, Cursor, Claude Code, or another coding agent. It can read the index first, then load the protocol, events, and samples it needs.

llms.txt

Compact index for discovering the documentation

Open

llms-full.txt

Complete context for a one-shot import

Open

What is in the skill

.agents/skills/tiktok-live-studio-local-data-access/
├── SKILL.md
├── agents/openai.yaml
├── assets/gifts.json
└── references/
    ├── protocol.md
    ├── events.md
    ├── gift-triggers.md
    ├── stack-adaptation.md
    └── verification.md

The skill is protocol-first rather than sample-first. It tells an agent how to inspect the developer's existing project and adapt the same connection lifecycle to JavaScript, TypeScript, Electron, H5, Unity, .NET, Java, Kotlin, Python, Go, Rust, C++, Unreal Engine, or another compatible local runtime.

What the agent learns

The skill covers:

• same-machine and loopback feasibility checks;
• port discovery and strict SERVER_HELLO validation;
• secure AUTH ordering and credential handling;
• event DTOs, optional fields, deduplication, and plain-text safety;
• gift catalog lookup, ID-based effect rules, and combo-safe triggering;
• connection ownership, cancellation, backoff, and reason 510 handling;
• stack-specific lifecycle and threading adaptation;
• mock-server tests, security review, and live smoke-test boundaries.

It also tells the agent what not to do: expose the service remotely, send credentials before verifying the endpoint, invent a JSON heartbeat, embed secrets in browser assets, or force the project into a different framework.

Use it in this repository

Codex, Cursor 2.4+, OpenCode, GitHub Copilot, and Gemini CLI can discover project skills from .agents/skills/ in supported versions. Invoke it explicitly when desired:

Use $tiktok-live-studio-local-data-access to integrate live.gift into this project.

Invocation syntax varies by agent. A tool-independent prompt is:

Read .agents/skills/tiktok-live-studio-local-data-access/SKILL.md and all references it requires, then use that workflow to integrate LIVE Studio local events into this project without changing the existing technology stack.

That explicit form also works for agents that do not automatically scan .agents/skills/.

Example requests

Use the LIVE Studio Local Data Access skill to add like and gift events to this Unity game. Reuse its existing WebSocket package.

Review this Python client against the LIVE Studio protocol. Focus on hello validation, authentication order, reconnect behavior, and secret logging.

Diagnose why this Electron overlay authenticates but stops receiving chat events after LIVE Studio restarts.

Keep the skill current

The public documentation remains the source of truth. When the endpoint, protocol version, message fields, event schemas, or retry guidance changes:

1. update the English and Chinese documentation;
2. update the matching files under the skill's references/ directory;
3. validate the skill and build the VitePress site;
4. redistribute the whole skill directory to projects that installed a copy.

The current skill targets protocol version 1.0.0.