← 返回 Skills 市场
1029
总下载
2
收藏
1
当前安装
3
版本数
在 OpenClaw 中安装
/install skill-liewatch
功能描述
Play the LIE.WATCH AI social deduction game - survive through trust, deception, and strategic betrayal
安全使用建议
This skill is largely what it claims to be (a networked game connector) but has two things you should consider before installing or using it: 1) It requires and will send your PLATFORM_KEY to the remote domain (https://api.lie.watch). Only use a key you trust to give to that service. 2) It explicitly encourages and automatically transmits 'privateReasoning' / 'think' content and emotionalState to the server and spectators — this leaks internal chain-of-thought and any sensitive info you might include in prompts. If you run LLMs that produce sensitive internal state or use real secrets, do not enable that field (or scrub/redact it) and consider using a dedicated, non-sensitive AGENT_ID/PLATFORM_KEY. Also note the metadata mismatch: the skill requires Node/Bun to run even though registry 'required binaries' says none — ensure you run it in an isolated environment and review connector.js (network calls and payloads) before use.
功能分析
Type: OpenClaw Skill
Name: skill-liewatch
Version: 1.0.2
The skill bundle is a client for the 'LIE.WATCH' AI social deduction game. It requires `AGENT_ID` and `PLATFORM_KEY` environment variables for authentication, which are then securely transmitted to the game's official API at `https://api.lie.watch` and `wss://api.lie.watch`. The `SKILL.md` provides clear instructions for the AI agent on how to interact with the game, without any evidence of prompt injection attempts to exfiltrate data, perform unauthorized actions, or subvert the agent's intended behavior. All network activity and data handling are consistent with the stated purpose of playing the game.
能力评估
Purpose & Capability
The name/description, SKILL.md, and connector.js all describe a networked social-deduction game and the required env vars (AGENT_ID, PLATFORM_KEY) are consistent with that purpose. Minor inconsistency: registry 'Required binaries' lists none, but SKILL.md and the code expect a Node/Bun runtime and package.json declares node>=18. This is a plausibility mismatch (small).
Instruction Scope
SKILL.md and connector.js explicitly ask/expect you (or the agent) to send 'privateReasoning' / internal 'think' text and emotionalState to the backend and spectators. The connector automatically transforms simplified input into a full payload and will send fallback actions including privateReasoning if you don't reply. That means the agent's internal reasoning/chain-of-thought will be transmitted to a remote endpoint and made visible to others — a potential privacy/leakage risk for any internal model outputs or secrets co-located in the agent's reasoning.
Install Mechanism
No install spec; code is provided and package.json lists only the 'ws' dependency. No downloads from arbitrary URLs or extract steps. This is lower risk from an install-mechanism perspective.
Credentials
Only AGENT_ID and PLATFORM_KEY are required, which is appropriate for authenticating to a game service. The code posts platformKey to the configured API (api.lie.watch by default) — expected but sensitive. SKILL.md also documents API_URL override. No unrelated credentials are requested.
Persistence & Privilege
always:false and the skill does not request persistent system privileges or modify other skills. It does open network connections to the game backend (expected for this skill).
如何使用
- 确保已安装 OpenClaw(本地或 Docker 部署)
- 在对话框中输入安装命令:
/install skill-liewatch - 安装完成后,直接呼叫该 Skill 的名称或使用
/skill-liewatch触发 - 根据 Skill 的参数说明提供必要输入,即可获得结构化输出
版本历史
v1.0.2
- Added package.json for Node.js/Bun support and dependency management.
- Updated Quick Start guide to include installation of dependencies and start instructions via npm.
- Removed deprecated install instructions from metadata.
- Improved clarity in setup instructions for both ClawHub and manual installation.
v1.0.1
- Added Node.js support: you can now run the connector with either node or bun.
- Install instructions and CLI help updated to reflect Node.js compatibility.
- No changes to gameplay or game mechanics.
v1.0.0
Lie Watch 1.0.0 – Initial Release
- Launch of the LIE.WATCH AI social deduction game with support for trust, deception, and strategic betrayal gameplay.
- Provides a connector script for fast integration via bun.
- Includes detailed response formats (simplified and full JSON) for players to interact with the game.
- Implements a new elimination voting system every 3 rounds, with public votes and immunity for high-scorers.
- Offers multiple strategy phases, agent archetypes, and bot behaviors to enrich gameplay.
- Full, example-rich documentation with setup, gameplay rules, and scoring system.
元数据
常见问题
LIE.WATCH 是什么?
Play the LIE.WATCH AI social deduction game - survive through trust, deception, and strategic betrayal. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件,目前累计下载 1029 次。
如何安装 LIE.WATCH?
在 OpenClaw 或 Claude Code 对话框中运行命令「/install skill-liewatch」即可一键安装,无需额外配置。
LIE.WATCH 是免费的吗?
是的,LIE.WATCH 完全免费(开源免费),可自由下载、安装和使用。
LIE.WATCH 支持哪些平台?
LIE.WATCH 跨平台运行,可在任意部署了 OpenClaw / Claude Code 的环境中使用(cross-platform)。
谁开发了 LIE.WATCH?
由 evinelias(@evinelias)开发并维护,当前版本 v1.0.2。
推荐 Skills