What problem does it solve? Deep Research features in Gemini, ChatGPT, and Perplexity run for many minutes and may pause for plan approval or clarifying questions, so they cannot be completed in a single agent turn. This Skill splits the job into deterministic start, check, and reply commands that drive a logged-in browser, letting an agent kick off long multi-source research, poll it later, and retrieve the finished cited report as a saved Markdown file. ## Core Features & Use Cases - Multi-site Deep Research automation: Starts Deep Research on Gemini, ChatGPT, or Perplexity by enabling the mode in the composer UI, submitting the prompt, and auto-approving the research plan. - Polling-based completion detection: The check command judges completion via report-text stability and streaming state, then saves report.md plus meta.json to a timestamped downloads directory. - Clarifying-question handling: Reads the conversation's last message when a site asks questions before starting, and answers it with the reply command. - Use Case: Ask "How have small modular reactors progressed since 2023?" — the agent starts Gemini Deep Research, schedules a reminder to poll every 6-8 minutes, and delivers the synthesized, citation-backed report when it finishes. ## Quick Start Ask the agent to run a deep research on your question, for example: start a Gemini deep research on how small modular reactors have progressed since 2023 and bring me the cited report when it finishes.
