baoyu-danger-gemini-web: a skill that calls Gemini's web app for text and images

baoyu-danger-gemini-web lets your agent generate text and images through the Gemini web app using your Google login, not an API key.

Installs: 28.1K; on skills.sh
Stars: 27.2K; on GitHub
Files: 27; besides SKILL.md
Updated: Nov 2026; last repo commit

The skill runs a bundled TypeScript client that talks to the Gemini web app through a reverse-engineered API. The agent can send a text prompt, ask for an image saved to a file, pass reference images for vision input, and keep a multi-turn conversation with a session id. Models include gemini-3-pro by default, gemini-3-flash and gemini-3-flash-thinking.

It is called "danger" because SKILL.md says it uses a reverse-engineered API. It requires your consent first. On the first run it opens Chrome so you can sign in to Google and then caches your cookies. The README warns of no stability guarantees.

Good for

People who want Gemini image generation inside their agent without an API key
Developers who need a Gemini backend for other baoyu skills
Anyone who accepts the risks of an unofficial web API

Install baoyu-danger-gemini-web

Claude Code

Copy this prompt.

Install the baoyu-danger-gemini-web skill in Claude Code. Run this command:

npx skills add https://github.com/jimliu/baoyu-skills --skill baoyu-danger-gemini-web

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Generates images and text via reverse-engineered Gemini Web API. Supports text generation, image generation from prompts, reference images for vision input, and multi-turn conversations. Use when other skills need image generation backend, or when user requests "generate image with Gemini", "Gemini text generation", or needs vision-capable AI generation.

Sections in SKILL.md

User Input Tools
Script Directory
Consent Check (REQUIRED)
Preferences (EXTEND.md)
Usage
Options
Models
Authentication
Environment Variables
Sessions
Extension Support

Key facts

Author: Jim Liu (JimLiu/baoyu-skills)
Repo: JimLiu/baoyu-skills, path skills/baoyu-danger-gemini-web
Installs on skills.sh: 28,074 (as of Nov 2, 2026)
GitHub stars: 27,159 (shared repo, Nov 2, 2026)
License: MIT unless otherwise noted
Version in SKILL.md: 1.56.2
Last repo commit: Nov 1, 2026
skills.sh audits: Gen Agent Trust Hub: warn. Socket: warn. Snyk: pass.
Files: 28 including SKILL.md: 25 TypeScript files plus package.json and bun.lock under scripts/
Needs: bun or npx, Chrome (or Chromium, Edge), a Google account

What it includes

Asks you to accept the unofficial-API disclaimer first
Checks consent.json in the gemini-web folder under your user data directory and creates it only after you accept.
Write · Consent check
main.ts "prompt" or --prompt, --promptfiles
Models set with --model: gemini-3-pro (default), gemini-3-flash, gemini-3-flash-thinking, gemini-3.1-pro-preview.
Runs code · Text generation
Generate an image
Saves to generated.png unless you give a path.
Write · --image [path]
Reference images for vision input
Send images with the prompt, for description or variations.
Network · --reference
Multi-turn conversations
Session files are stored in the data directory under sessions/<id>.json.
Write · --sessionId, --list-sessions
Refresh cookies and exit
Opens or reuses Chrome so you can sign in to Google.
Runs code · --login
Output and storage options
--profile-dir forces a dedicated Chrome profile.
--json, --cookie-path, --profile-dir

From the official docs

Permissions

High risk

It uses your Google login cookies with an unofficial web API and runs a TypeScript script that drives Chrome. The README gives no stability guarantees and says to use it at your own risk, so use a spare account.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It uses an unofficial API with your Google login
The README says this is reverse-engineered, unofficial access through browser cookies with no stability guarantees. Use a spare Google account. github.com
It can reuse a running Chrome session
Without an explicit profile, cookie refresh may attach to an already-running Chrome debugging session. Set --profile-dir or GEMINI_WEB_CHROME_PROFILE_DIR to force a dedicated profile. github.com
It runs third-party code
Scripts run with bun, or `npx -y bun` if bun is missing. scripts/package.json depends on the npm package baoyu-chrome-cdp. github.com
Your prompts and images go to Google
Text and reference images are sent to the Gemini web app, and sessions are saved as JSON files on your disk. github.com

Technical details

Talks to	gemini.google.com and accounts.google.com. A proxy you set with HTTP_PROXY or HTTPS_PROXY
Sign-in and keys	Cookies are cached in a cookie file in the data directory. SKILL.md says cookie refresh may reuse an already-running local Chrome or Chromium debugging session.
Scopes	Google login cookies

Questions

What does baoyu-danger-gemini-web do?

It lets your agent generate text and images through the Gemini web app, with optional reference images and saved multi-turn sessions, using a TypeScript client that signs in with your Google cookies.

How do I install it in Claude Code?

Use the install command in the box above. The README also lists a plugin marketplace route that brings every skill in the repo, and advises installing only the skills you need.

Does it need a Gemini API key?

No. It uses your Google login through Chrome instead. The first run opens a browser to sign in and then caches the cookies.

Is it safe?

It is high risk by the author's own wording: an unofficial API, Google session cookies and a script that drives Chrome. Use a spare account, read the disclaimer it shows and keep the cookie file private.

Which models can it use?

gemini-3-pro (default), gemini-3-flash, gemini-3-flash-thinking and gemini-3.1-pro-preview, set with --model.

Can I use a proxy?

Yes. The README shows setting HTTP_PROXY and HTTPS_PROXY inline in front of the command if you need a proxy to reach Google.

Why is there a consent prompt?

SKILL.md requires you to accept a disclaimer for reverse-engineered API use before first use. Your acceptance is stored in a consent.json file.

Information

updated Nov 2, 2026

Repository: jimliu/baoyu-skills
Folder: skills/baoyu-danger-gemini-web

jimliu/baoyu-skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2

SKILL.md · github.com
baoyu-skills README · github.com