What problem does it solve? Literature reviews built from quick searches often miss key works, cite papers that do not exist, and read as paper-by-paper summaries rather than argued surveys. This Skill runs a rigorous, multi-phase research investigation that freezes research questions first, searches from multiple adversarial perspectives, verifies every citation, and synthesizes evidence into a structured survey report. ## Core Features & Use Cases - Research brief freezing: Converts a vague topic into two or three specific, answerable research questions before any searching begins. - Multi-perspective scouting: Searches from mainstream, critic, adjacent-field, methodology, and policy perspectives, with optional parallel sub-agents, plus blind-spot checks for coverage gaps. - Citation verification protocol: Every reference passes a two-step existence and quote-accuracy check with a five-grade verdict; unverifiable citations are never used. - MECE synthesis and quality gates: Clusters works into a mutually exclusive taxonomy, enforces in-sentence cross-comparison, and iterates through six internal quality gates plus a self-adversarial review. - Use Case: Ask for a survey of retrieval-augmented generation for question answering, and receive a full Markdown survey with abstract, methodology, taxonomy, comparison tables, open problems, and a verified reference list. ## Quick Start Ask the agent to run a deep research survey on your topic, for example: survey the current state of CRISPR clinical therapies and map the key works, debates, and open problems.
