Generate Comprehensive Research Reports

Gemini Deep Research Skill runs autonomous research tasks that plan, search, read, and synthesize information into comprehensive, cited reports via Google's Gemini API. A CLI script (scripts/research.py) accepts a --query and supports a custom output --format (e.g. "1. Executive Summary\n2. Comparison Table\n3. Recommendations"), can --stream progress in real time or run --no-wait to start a task asynchronously without blocking, and lets you --status or --wait on a running interaction_id, --continue a prior research thread with a follow-up query, or --list recent research runs. Output defaults to a human-readable markdown report, with --json for structured programmatic output and --raw for the unprocessed API response.

python3 scripts/research.py --query "Research the history of Kubernetes"

Each task takes roughly 2-10 minutes and costs $2-5 depending on complexity, consuming approximately 250k-900k input tokens and 60k-80k output tokens - cost and latency the assistant should surface to the user up front rather than let them discover mid-wait.

Best suited to market analysis and competitive landscaping, technical literature reviews, due diligence research, historical research and timelines, and comparative analysis of frameworks, products, or technologies - any task needing a detailed, cited research report. Not for a quick factual lookup: the 2-10 minute runtime and $2-5 per-task cost aren't justified when a direct answer would do.

Requires Python 3.8+, the httpx dependency (pip install -r requirements.txt), and a GEMINI_API_KEY environment variable obtained from Google AI Studio, set via export or a .env file in the skill directory. Given a research query, it returns a formatted report (or JSON/raw variants) plus an interaction_id for status checks, waiting, or follow-up continuation. Exit codes signal outcome: 0 for success, 1 for an API error, config issue, or timeout, and 130 if cancelled by the user via Ctrl+C.

The documented workflow makes the division of responsibility explicit: the assistant runs --query, informs the user of the estimated 2-10 minute runtime up front, monitors progress with --stream or polls with --status, returns the formatted result, and uses --continue for any follow-up questions on the same research thread - the Gemini API itself does the planning, searching, reading, and synthesis behind that CLI surface, authenticated solely via GEMINI_API_KEY.

Analysts, researchers, and engineers who need autonomous, multi-step research synthesized into a cited report - particularly for market/competitive analysis, literature reviews, or due diligence - without manually running and stitching together multiple search-and-read passes themselves.