Google Web Search

Enables grounded question answering by automatically executing the Google Search tool within Gemini models. Use when the required information is recent (post knowledge cutoff) or requires verifiable citation.

Security Scan

The skill's code, instructions, and required credential (GEMINI_API_KEY) are consistent with its stated purpose of using Gemini's google_search grounding tool; nothing requests unrelated secrets or installs arbitrary remote code.

Purpose & Capability

Name/description match the implementation: the skill uses google-genai to call Gemini's google_search grounding tool. The only required credential is GEMINI_API_KEY and the listed Python dependencies (google-genai, pydantic-settings) are exactly what you'd expect for this functionality.

Instruction Scope

SKILL.md instructs installing the Python deps and invoking scripts/example.py. The runtime code reads GEMINI_API_KEY (and optionally GEMINI_MODEL) from the environment. The pydantic BaseSettings config references an env_file '.env', so the skill may also load variables from a local .env file in the working directory — this is not malicious but worth noting because it can pull secrets from .env files if present.

Install Mechanism

No custom download or executable installer; install is just 'pip install -r requirements.txt' for standard PyPI packages. The declared dependencies are reasonable and traceable (google-genai, pydantic-settings).

Credentials

Only GEMINI_API_KEY is required (GEMINI_MODEL optional). No unrelated credentials, system config paths, or broad secret requests are present. The use of .env means local environment files can supply the key, which is expected for this kind of skill.

Persistence & Privilege

Skill is not forced-always and does not request elevated platform privileges. It can be invoked autonomously (default), which is normal for skills; there is no evidence it modifies other skills or system-wide settings.

Assessment

This skill appears to do what it claims: call Gemini's google_search grounding tool. Before installing, 1) ensure you supply a Gemini API key with minimal permissions (rotate/restrict the key where possible), 2) run the pip install in an isolated Python environment (venv/container) to avoid contaminating system packages, 3) be aware the code will read environment variables and will also load a local .env file if present (remove or audit any .env files that contain unrelated secrets), and 4) review network/policy requirements for allowing outbound calls to Google's API from the environment where you run this skill.

Overview

This skill provides the capability to perform real-time web searches via the Gemini API's google_search grounding tool. It is designed to fetch the most current information available on the web to provide grounded, citable answers to user queries.

Key Features:

• Real-time web search via Gemini API
• Grounded responses with verifiable citations
• Configurable model selection
• Simple Python API

Usage

This skill exposes the Gemini API's google_search tool. It should be used when the user asks for real-time information, recent events, or requests verifiable citations.

Execution Context

The core logic is in scripts/example.py. This script requires the following environment variables:

• GEMINI_API_KEY (required): Your Gemini API key
• GEMINI_MODEL (optional): Model to use (default: gemini-2.5-flash-lite)

Supported Models:

• gemini-2.5-flash-lite (default) - Fast and cost-effective
• gemini-3-flash-preview - Latest flash model
• gemini-3-pro-preview - More capable, slower
• gemini-2.5-flash-lite-preview-09-2025 - Specific version

Python Tool Implementation Pattern

When integrating this skill into a larger workflow, the helper script should be executed in an environment where the google-genai library is available and the GEMINI_API_KEY is exposed.

Example Python invocation structure:

from skills.google-web-search.scripts.example import get_grounded_response

# Basic usage (uses default model):
prompt = "What is the latest market trend?"
response_text = get_grounded_response(prompt)
print(response_text)

# Using a specific model:
response_text = get_grounded_response(prompt, model="gemini-3-pro-preview")
print(response_text)

# Or set via environment variable:
import os
os.environ["GEMINI_MODEL"] = "gemini-3-flash-preview"
response_text = get_grounded_response(prompt)
print(response_text)

Troubleshooting

If the script fails:

1. Missing API Key: Ensure GEMINI_API_KEY is set in the execution environment.
2. Library Missing: Verify that the google-genai library is installed (pip install google-generativeai).
3. API Limits: Check the API usage limits on the Google AI Studio dashboard.
4. Invalid Model: If you set GEMINI_MODEL, ensure it's a valid Gemini model name.
5. Model Not Supporting Grounding: Some models may not support the google_search tool. Use flash or pro variants.