Gemini Deep Research

Perform complex, long-running research tasks using Gemini Deep Research Agent. Use when asked to research topics requiring multi-source synthesis, competitive analysis, market research, or comprehensive technical investigations that benefit from systematic web search and analysis.

Security Scan

The skill's code, runtime instructions, and requested credential (GEMINI_API_KEY) are consistent with a Gemini Deep Research client; nothing in the files indicates unrelated access or covert exfiltration.

Purpose & Capability

Name/description match the implementation: the script talks to the generativelanguage.googleapis.com interactions API with the declared agent and uses an x-goog-api-key header. The single required credential (GEMINI_API_KEY) is appropriate for this purpose.

Instruction Scope

SKILL.md and the script limit actions to creating/polling Gemini interactions, extracting model output, optional file-search tool usage (via fileSearchStore name sent to the API), and writing results to local timestamped files. The instructions do not read arbitrary local files or access other environment variables.

Install Mechanism

There is no install spec (instruction-only), which minimizes disk writes. However, a Python script is included that depends on a Python runtime and the 'requests' library; those dependencies are not declared in the skill metadata or install instructions. This is not a security issue but is an operational omission the user should be aware of.

Credentials

Only GEMINI_API_KEY is required and used as documented (x-goog-api-key). No other credentials, secrets, or unrelated environment variables are requested or accessed.

Persistence & Privilege

The skill is not always-enabled and does not request elevated platform privileges or modify other skills. It writes output files to the specified output directory (default is current directory), which is expected behavior for a CLI research tool.

Assessment

This skill appears to do what it says: it uses a Gemini API key to start and poll Deep Research agent interactions and saves reports locally. Before installing/running, confirm: (1) you are comfortable providing a Gemini API key—this key allows access to your Google AI Studio quota and may incur costs; (2) the key's scope and permissions (and whether it can access any internal file-search stores) meet your security policies; (3) you have a Python runtime and the 'requests' package available (the skill does not declare an installer); and (4) review the script if you plan to run it with sensitive inputs (it will send queries and optional fileSearchStore names to Google’s API and will write outputs to disk). If you need stronger guarantees, request a version that includes a manifest specifying runtime and dependency installation steps and/or remove the file-search option if you do not want the agent to reference private stores.

Use Gemini's Deep Research Agent to perform complex, long-running context gathering and synthesis tasks.

Prerequisites

• GEMINI_API_KEY environment variable (from Google AI Studio)
• Note: This does NOT work with Antigravity OAuth tokens. Requires a direct Gemini API key.

How It Works

Deep Research is an agent that:

1. Breaks down complex queries into sub-questions
2. Searches the web systematically
3. Synthesizes findings into comprehensive reports
4. Provides streaming progress updates

Usage

Basic Research

scripts/deep_research.py --query "Research the history of Google TPUs"

Custom Output Format

scripts/deep_research.py --query "Research the competitive landscape of EV batteries" \
  --format "1. Executive Summary\n2. Key Players (include data table)\n3. Supply Chain Risks"

With File Search (optional)

scripts/deep_research.py --query "Compare our 2025 fiscal year report against current public web news" \
  --file-search-store "fileSearchStores/my-store-name"

Stream Progress

scripts/deep_research.py --query "Your research topic" --stream

Output

The script saves results to timestamped files:

• deep-research-YYYY-MM-DD-HH-MM-SS.md - Final report in markdown
• deep-research-YYYY-MM-DD-HH-MM-SS.json - Full interaction metadata

API Details

• Endpoint: https://generativelanguage.googleapis.com/v1beta/interactions
• Agent: deep-research-pro-preview-12-2025
• Auth: x-goog-api-key header (NOT OAuth Bearer token)

Limitations

• Requires Gemini API key (get from Google AI Studio)
• Does NOT work with Antigravity OAuth authentication
• Long-running tasks (minutes to hours depending on complexity)
• May incur API costs depending on your quota