What problem does it solve? Research write-ups often contain unverified claims, dead citations, and hallucinated sources. This Skill produces research documents where every load-bearing claim is fetched live, cross-checked, tagged by evidence tier, and recorded in a provenance ledger. ## Core Features & Use Cases - Grounded research workflow: Assembles a prompt from a bundled template enforcing Chain-of-Verification, multi-framing search fan-out, and a retrieval floor so no section ends without a fetched source. - Adversarial review and citation validation: Runs five critic perspectives (conventional, contrarian, historical, adjacent, skeptic) and fetches every citation, classifying access failures such as 404s, paywalls, and DNS errors. - Provenance ledger and claim tiers: Tags claims as [VERIFIABLE], [HEURISTIC], [INFERENCE], or [NO SOURCE], with a claim-to-quote-to-verdict ledger for every verifiable statement. - Use Case: Ask for a scoped write-up like "research vector database consistency models"; the Skill scaffolds docs/research/<slug>.md from the bundled stub, researches with live fetches, and delivers a document with footnoted sources and a full audit trail. ## Quick Start Ask the agent to research a topic and write it up, for example: use research-doc to investigate current approaches to LLM evaluation and produce a cited research document.
