What problem does it solve? Single-agent web research is slow, shallow, and prone to accepting unverified claims. This Skill orchestrates a swarm of parallel research subagents that decompose a topic, run massive numbers of web searches, scrape full pages, and stress-test findings before delivering a consolidated report. ## Core Features & Use Cases - Tiered Depth Modes: Choose from 6 modes (Very Fast, Fast, Lite, Deeper, Ultra, and a 5-tier fractal mode scaling up to ~550 subagents) with defined search budgets per mode. - Verification & Safety Controls: Includes 102% overthink verification, instant premise correction, mandatory full-page scraping, a strict 2-minute hard timeout, and execution telemetry logging. - Structured Deliverables: Produces a plain-English research report with executive summary, fact-check alerts, comparison tables, edge-case analysis, and a verified source directory, plus post-research swarm lifecycle management. - Use Case: Ask for a deep dive on a technical topic like GPU benchmark comparisons; the swarm fans out across dozens of searches, scrapes documentation and benchmark tables, verifies claims, and returns a single consolidated report within two minutes. ## Quick Start Run /deep-research on the topic of your choice and select a depth mode when prompted.
