← 返回 Skills 市场
1915
总下载
3
收藏
2
当前安装
1
版本数
在 OpenClaw 中安装
/install tokenguard
功能描述
API cost guardian for AI agents. Track spending, enforce limits, prevent runaway costs. Essential for any agent making paid API calls.
安全使用建议
TokenGuard appears coherent and limited to local budgeting: it reads/writes JSON under ~/.tokenguard and exposes check/log commands you can call from agents or scripts. Before installing, review the complete scripts/tokenguard.py file (the submitted copy was truncated), verify the repository/homepage match the distributed file, and confirm there are no hidden network calls or unexpected file accesses. Use a dedicated non-sensitive directory for TOKENGUARD_DIR, keep file permissions minimal, and be aware that an automated agent with permission to run the script could set the 'override' flag to bypass limits — ensure your agent's policy/control flow prevents unintended overrides.
功能分析
Type: OpenClaw Skill
Name: tokenguard
Version: 1.0.0
The OpenClaw AgentSkills skill bundle 'tokenguard' is a benign utility designed to track and limit API spending for AI agents. The `SKILL.md` documentation clearly outlines its purpose and usage, without any prompt injection attempts or instructions for the agent to perform unauthorized actions. The core script `scripts/tokenguard.py` implements the described functionality, storing its data (limit, session, override flag) in a dedicated directory (`~/.tokenguard` by default) within the user's home directory. It uses standard Python libraries, reads only its own configuration environment variables, and does not exhibit any signs of data exfiltration, malicious execution, persistence mechanisms, or obfuscation. All file operations and capabilities are strictly aligned with its stated purpose of API cost management.
能力评估
Purpose & Capability
The name/description (API cost guardian) aligns with the included code and SKILL.md: the script implements set/check/log/history/reset/extend/override and stores state under ~/.tokenguard (or TOKENGUARD_DIR). No unrelated services, credentials, or system subsystems are requested.
Instruction Scope
Runtime instructions are scoped to running the local Python script and storing/retrieving JSON in its own directory. Integration guidance uses subprocess to call the script (expected). The SKILL.md does not instruct reading other system config, secrets, or sending data to external endpoints.
Install Mechanism
There is no automatic install action (instruction-only). Manual install steps simply copy the SKILL.md and the script; that is low-risk. The package declares a GitHub repository but does not pull remote code during install.
Credentials
No required environment variables or credentials are declared; SKILL.md documents optional env vars (TOKENGUARD_DIR, TOKENGUARD_DEFAULT_LIMIT, TOKENGUARD_WARNING_PCT) used only to configure storage and defaults. No secrets or external API keys are requested.
Persistence & Privilege
The skill writes only to its own directory (~/.tokenguard by default) and does not request always:true or modify other skills. It can be invoked autonomously (platform default), which is normal for skills; note that the 'override' feature allows a one-time bypass and could be invoked programmatically if an agent is permitted to run commands.
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 tokenguard - 安装完成后，直接呼叫该 Skill 的名称或使用
/tokenguard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
v1.0.0
Initial release
元数据
常见问题
TokenGuard 是什么？
API cost guardian for AI agents. Track spending, enforce limits, prevent runaway costs. Essential for any agent making paid API calls. 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 1915 次。
如何安装 TokenGuard？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install tokenguard」即可一键安装，无需额外配置。
TokenGuard 是免费的吗？
是的，TokenGuard 完全免费（开源免费），可自由下载、安装和使用。
TokenGuard 支持哪些平台？
TokenGuard 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 TokenGuard？
由 G0HEAD（@g0head）开发并维护，当前版本 v1.0.0。
推荐 Skills