Skills baoyu-danger-gemini-web

📦

baoyu-danger-gemini-web

Name: baoyu-danger-gemini-web
Author: jimliu

v1.56.2 Content revision r2 High Risk 🌐 Network access📁 Filesystem access⚙️ External commands🔑 Env variables

Generate Gemini Text and Images

Connects AI workflows to Gemini Web generation without a separate API integration. Produces text, images, vision responses, and multi-turn results from a command-line interface.

Supports: Claude Codex Code(CC)

⚠️ 38 Poor

Stop for confirmation before installing.

Review the plan and obtain explicit user consent before changing files.

Install with my Agent

Copy this request to your Agent. It includes the canonical Skill page and manifest.

Agent request

Review the Skillstore skill "baoyu-danger-gemini-web" from https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web.md and its manifest at https://skillstore.io/api/skills/jimliu-baoyu-danger-gemini-web/manifest. Verify the artifact. Stop and obtain explicit user consent before installing or changing files.

Your Agent should still show its plan and request any confirmation required by the security policy.

Agent-readable resources

Use these links when an AI agent, crawler, or script needs clean context instead of reading the full page.

Markdown detail GET /skills/jimliu-baoyu-danger-gemini-web.md Signed manifest GET /api/skills/jimliu-baoyu-danger-gemini-web/manifest Signed lockfile GET /api/skills/jimliu-baoyu-danger-gemini-web/lockfile

Test it

Using "baoyu-danger-gemini-web". Create a square editorial illustration of a quiet mountain research station at dawn.

Expected outcome:

A generated image is saved to the selected output directory, with the returned title and image metadata available to the caller.

Using "baoyu-danger-gemini-web". Describe the attached product photo and suggest three improvements for its presentation.

Expected outcome:

The response identifies the product, framing, lighting, and visible distractions.
It suggests three concrete changes to improve clarity and presentation.

Security Audit

High Risk

v5 • 8/7/2026 Open versioned report

Most low and medium static matches are false positives from documented Google endpoints, relative imports, template literals, configuration reads, and normal session storage. Confirmed risks include cached authentication-cookie access, shell process execution for WSL detection, browser cookie reuse, and an output filename path escape; the skill should not be published until these are addressed.

Files scanned

2,970

Lines analyzed

Review items

False positives ignored

Confirmed security concerns (4)

High

scripts/gemini-webapi/utils/cookie-file.ts:46

Certificate/key files

return Object.keys(out).length > 0 ? out : null;

The code reads and returns persisted authentication cookie values from a local file. These cookies can authorize Google sessions, so compromise of the file exposes credentials even though the scanner label is imprecise.

High

scripts/gemini-webapi/utils/cookie-file.ts:55

Certificate/key files

return Object.keys(out).length > 0 ? out : null;

High

scripts/gemini-webapi/utils/load-browser-cookies.ts:105-119 scripts/gemini-webapi/utils/load-browser-cookies.ts:150-158

Authenticated Browser Session Reuse

The client discovers an existing Chrome debugging session, retrieves Google cookies through CDP, and persists them for authenticated Gemini requests. This can grant the skill access to an active Google account session.

The source directly discovers a running browser, calls Network.getCookies for Google domains, and returns the cookie values for later authenticated requests.

High

scripts/gemini-webapi/types/image.ts:73-78

Output Filename Path Escape

Image.save joins the resolved output directory with a filename without constraining the filename to a basename. A caller-supplied filename containing traversal segments could write outside the intended directory.

The destination is built with path.join from a caller-influenced filename, and no basename check or resolved-path containment check is present before writeFile.

Capability review items (8)

These are real local capabilities that may be expected for this skill, so they require review but are not counted as confirmed malicious behavior.

High

scripts/gemini-webapi/utils/paths.ts:1

Node.js child_process module

import { execSync } from 'node:child_process';

The code invokes a system process through execSync for WSL and Windows path detection. The wslpath command interpolates command output into a shell string, creating a genuine process-execution and injection surface.

High

scripts/gemini-webapi/utils/paths.ts:38

Synchronous process execution

const raw = execSync('cmd.exe /C "echo %USERPROFILE%"', { encoding: 'utf-8', timeout: 5000 }).trim()

High

scripts/gemini-webapi/utils/paths.ts:39

Synchronous process execution

_wslHome = execSync(`wslpath -u "${raw}"`, { encoding: 'utf-8', timeout: 5000 }).trim() || null;

High

scripts/gemini-webapi/utils/paths.ts:38

Windows cmd.exe

const raw = execSync('cmd.exe /C "echo %USERPROFILE%"', { encoding: 'utf-8', timeout: 5000 }).trim()

Medium

scripts/gemini-webapi/utils/get-access-token.ts:44

Node.js fs operations

const v = fs.readFileSync(p, 'utf8').trim();

The code reads cached authentication material from local files and enumerates cached token files. These values can authorize a Google session, so local credential access is a real security risk.

Medium

scripts/gemini-webapi/utils/get-access-token.ts:55

Node.js fs operations

for (const f of fs.readdirSync(dir)) {

The code reads cached authentication material from local files and enumerates cached token files. These values can authorize a Google session, so local credential access is a real security risk.

Medium

scripts/gemini-webapi/utils/get-access-token.ts:44

Synchronous file operations

const v = fs.readFileSync(p, 'utf8').trim();

The code reads cached authentication material from local files and enumerates cached token files. These values can authorize a Google session, so local credential access is a real security risk.

Medium

scripts/gemini-webapi/utils/paths.ts:39

Ruby/shell backtick execution

_wslHome = execSync(`wslpath -u "${raw}"`, { encoding: 'utf-8', timeout: 5000 }).trim() || null;

The template literal is passed directly to execSync for shell execution, and it interpolates the result of another command into the wslpath string. This creates a genuine process-execution and injection surface.

Risk Factors

🌐 Network access (23)

📁 Filesystem access (34)

⚙️ External commands (50)

🔑 Env variables (26)

Audited by: codex View Audit History →

Share & cite this report

Share the versioned assessment report, neutral badge, embed card, and citations. Skillstore reports evidence without deciding whether this Skill is safe.

Open versioned report

Copy report link

https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web/audits/5?utm_source=security_passport&utm_medium=share&utm_campaign=versioned_report

Markdown badge

[![Skillstore security assessment](https://skillstore.io/badges/skills/jimliu-baoyu-danger-gemini-web/security.svg)](https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web?utm_source=security_passport_badge)

HTML badge

<a href="https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web?utm_source=security_passport_badge"><img src="https://skillstore.io/badges/skills/jimliu-baoyu-danger-gemini-web/security.svg" alt="Skillstore security assessment" loading="lazy"></a>

Embed card

<iframe src="https://skillstore.io/embed/skills/jimliu-baoyu-danger-gemini-web.html" title="Skillstore Security Assessment" sandbox="allow-popups allow-popups-to-escape-sandbox" loading="lazy" referrerpolicy="no-referrer" width="420" height="180"></iframe>

Academic citations (APA · BibTeX · CFF)

APA citation

jimliu. (2026). baoyu-danger-gemini-web security audit report (audit version 5) [Author version 1.56.2]. Skillstore. https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web/audits/5

BibTeX citation

@techreport{jimliu-jimliu-baoyu-danger-gemini-web-2026,
  author = {jimliu},
  title = {baoyu-danger-gemini-web security audit report (audit version 5)},
  institution = {Skillstore},
  year = {2026},
  number = {5},
  url = {https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web/audits/5},
  note = {Author version 1.56.2}
}

CITATION.cff

cff-version: 1.2.0
message: "If you use this Skill, cite its author and this versioned security audit report."
title: "baoyu-danger-gemini-web security audit report (audit version 5)"
version: "1.56.2"
type: report
authors:
  - name: "jimliu"
date-released: "2026-08-07"
url: "https://skillstore.io/skills/jimliu-baoyu-danger-gemini-web/audits/5"
identifiers:
  - type: other
    value: "skillstore:jimliu-baoyu-danger-gemini-web:audit:5"
    description: "Skillstore immutable audit report identifier"

Skillstore Score

Why this score Evidence Confidence: High

Architecture

100

Maintainability

Content

Community

Spec Compliance

What You Can Build

Prototype Visual Concepts

Generate concept images, variations, and visual references while exploring an early creative direction.

Run Vision Experiments

Send reference images with focused prompts to compare descriptions, transformations, and multimodal responses.

Add Gemini to Workflows

Call Gemini text or image generation from an existing automation workflow through a simple command-line interface.

Try These Prompts

Simple Text Request

Generate a concise explanation of [topic] for [audience]. Use clear headings and three practical examples.

Create an Image

Create a detailed image of [subject] in [style]. Use [lighting], [color direction], and a [composition] suitable for [use].

Analyze a Reference

Review the reference image and describe its subject, composition, lighting, and visual style. List three specific improvement options.

Continue a Session

Continue the current conversation. Use the previous constraints, then produce [requested result] with these checks: [check one], [check two].

Best Practices

Use a dedicated Google account and browser profile for this integration.
Review prompts, reference images, and upload destinations before each run.
Keep cached cookies and session files outside shared directories with restrictive permissions.

Avoid

Do not use this skill with a shared browser profile containing unrelated accounts.
Do not send confidential prompts or images without confirming the destination and retention expectations.
Do not rely on reverse-engineered endpoints for critical production guarantees.