← 返回 Skills 市场
Baoyu Danger Gemini Web
作者
Jim Liu 宝玉
· GitHub ↗
· v1.56.2
· MIT-0
1736
总下载
0
收藏
50
当前安装
3
版本数
在 OpenClaw 中安装
/install baoyu-danger-gemini-web
功能描述
Generates images and text via reverse-engineered Gemini Web API. Supports text generation, image generation from prompts, reference images for vision input,...
安全使用建议
Install only if you are comfortable letting this skill use a Google/Gemini web session. Prefer a dedicated Chrome profile and private data directory, avoid using it with a high-value Google account, do not pass sensitive reference files or prompts, and periodically delete its saved cookies and sessions if you no longer need them.
能力评估
Purpose & Capability
The stated purpose is Gemini text/image generation, and the code supports that, but it also obtains Google/Gemini browser cookies through Chrome DevTools, stores reusable session cookies, uploads reference files, saves chat sessions, and includes Gemini Gem management helpers. Those are mostly purpose-aligned, but they are high-impact account/session capabilities.
Instruction Scope
SKILL.md discloses reverse-engineered API use, browser auth, automatic cookie caching, existing Chrome debugging-session reuse, and session storage. However, the consent check is an instruction to the agent rather than enforced in the CLI, and the routing guidance is broad for a skill with sensitive credential access.
Install Mechanism
The artifact manifest declares bun/npx requirements but does not fully declare network, filesystem persistence, Chrome/CDP access, or sensitive cookie handling in the install-facing openclaw requires block. Package metadata does separately tag the skill as requiring sensitive credentials.
Credentials
Network calls to Google/Gemini and user-selected file uploads fit the stated Gemini Web purpose, but default fallback to a reusable local Chrome debugging session and broad Google cookie collection are more sensitive than a typical generation backend.
Persistence & Privilege
The code writes authentication cookies and refreshed __Secure-1PSIDTS values to local files, and --sessionId stores prompt/assistant transcripts. These behaviors are disclosed in part, but cookie storage is plaintext and auto-refresh extends the lifetime of local session material.
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 baoyu-danger-gemini-web - 安装完成后，直接呼叫该 Skill 的名称或使用
/baoyu-danger-gemini-web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
v1.56.2
Fix stale npm shared package dependency pins for baoyu-md and baoyu-chrome-cdp 0.1.1.
v1.56.1
Version 1.56.1
- Added: `version` and `metadata` fields to SKILL.md for improved compatibility and discoverability.
- Updated: User input tool selection guidelines for better integration across diverse runtimes.
- Enhanced: Script instructions now support both `bun` and `npx` fallback, with runtime variable `${BUN_X}`.
- Improved: Preferences section now supports XDG config spec for EXTEND.md location; updated priority/search order.
- Clarified: Authentication instructions regarding Chrome profile reuse and session handling.
- Added: New scripts - `scripts/bun.lock`, `scripts/gemini-webapi/client.test.ts`, `scripts/package.json` for reproducible builds and improved testability.
v0.1.0
Initial release of baoyu-danger-gemini-web.
- Provides text and image generation via the reverse-engineered Gemini Web API, including multi-turn conversations and vision input through reference images.
- Requires explicit user consent before first use; handles consent storage and checks across macOS, Linux, and Windows.
- Supports configuration via EXTEND.md at both project and user levels for model selection, proxy settings, and data directories.
- Includes CLI support for text/image generation, session handling, JSON output, and browser-based authentication.
- Supports several Gemini-3 models, with flexible authentication, session, and environment variable options for customization.
元数据
常见问题
Baoyu Danger Gemini Web 是什么？
Generates images and text via reverse-engineered Gemini Web API. Supports text generation, image generation from prompts, reference images for vision input,... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 1736 次。
如何安装 Baoyu Danger Gemini Web？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install baoyu-danger-gemini-web」即可一键安装，无需额外配置。
Baoyu Danger Gemini Web 是免费的吗？
是的，Baoyu Danger Gemini Web 完全免费，采用 MIT-0 许可证，可自由下载、安装和使用。
Baoyu Danger Gemini Web 支持哪些平台？
Baoyu Danger Gemini Web 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 Baoyu Danger Gemini Web？
由 Jim Liu 宝玉（@jimliu）开发并维护，当前版本 v1.56.2。
推荐 Skills