What problem does it solve? Web research often produces unverified claims, stale sources, and single-source assertions that cannot be trusted. This Skill runs a multi-phase research swarm where independent Fact-Checkers and Devil's Advocate agents cross-examine every claim before it reaches the final report, so only corroborated findings survive. ## Core Features & Use Cases - Adversarial Verification Pipeline: Parallel researchers gather claims, then Fact-Checkers independently verify them and Devil's Advocates actively try to disprove confirmed findings. - Scalable Depth Modes: Choose from Swift (5 agents), Strike (10), Storm (15), or Siege (20) depending on how much verification firepower the topic needs. - Dispute Resolution & Confidence Scoring: Contradicted claims go through Pro/Con debate with a Judge, and every claim receives a confidence tier from Verified to Debunked. - Use Case: Ask for verified research on a fast-moving topic like a new AI framework; the swarm decomposes it into angles, floods searches, stress-tests every claim, and delivers a report separating verified facts from disputed and debunked ones. ## Quick Start Run /boost-research on the topic "current state of local LLM inference frameworks" and select Strike mode to get an adversarially verified research report.
