What problem does it solve? Conducting rigorous multi-source research manually means juggling dozens of searches, tracking which claims came from which sources, and verifying citations — a process prone to hallucinated references and lost evidence. This Skill automates the entire pipeline, producing citation-backed reports with an auditable evidence trail persisted to disk. ## Core Features & Use Cases - 8-Phase Research Pipeline: Scope, plan, retrieve (parallel searches plus sub-agents), triangulate, synthesize, critique with loop-back, refine, and package — with four depth modes (quick, standard, deep, ultradeep) from 2 to 45 minutes. - Evidence & Claim Persistence: Sources, evidence quotes, and atomic claims are stored in append-only JSONL files (sources.jsonl, evidence.jsonl, claims.jsonl) with stable SHA-256 identities, surviving context compaction and continuation agents. - Multi-Format Output: Generates Markdown, McKinsey-style HTML, and PDF reports with automated validation (structure checks, citation verification, hallucination detection) and auto-continuation for reports over 18,000 words. - Use Case: Ask for a deep comparison of PostgreSQL vs Supabase for your stack, and receive a verified report with 10+ sources, 3+ citations per major claim, and a complete bibliography. ## Quick Start Ask the agent to do deep research on the current state of quantum computing in standard mode and save the report to the project docs folder.
