Skill UI

This advanced skill systematically verifies the novelty of a proposed academic method or research idea against the latest scientific literature. It performs multi-source searches across databases like arXiv, Google Scholar, and major conference proceedings (ICML, NeurIPS). Using a multi-step process, it extracts core claims, searches multiple sources, performs cross-model verification via LLM, and outputs a structured Novelty Report, detailing overlap, prior work, and overall recommendation.

This critical skill serves as a gate between completed experiments and final research conclusions. It systematically evaluates collected experimental results (e.g., W&B metrics, logs) against predefined research claims. It performs deterministic evidence pre-checks to verify if cited data points exist, and utilizes an advanced AI model (Codex) to judge whether the results support, contradict, or lack evidence for the intended claims, guiding the next action (pivot, supplement, or confirm).

A persistent, structured knowledge base designed to accumulate and connect all research artifacts—including papers, ideas, experiments, and claims—across the entire research lifecycle. Inspired by LLM Wiki patterns, it builds a comprehensive, evolving graph of knowledge by tracking typed relationships (e.g., supports, contradicts, extends), ensuring that findings are compounded and never re-derived. Ideal for complex, multi-stage scientific projects.

This powerful tool performs real-time, multi-backend research lookup by intelligently routing queries to the optimal source. It supports fast web searches, deep multi-source synthesis (Chat API), and scholarly paper searches (Perplexity). Use it to collect verifiable statistics, research market trends, verify scientific claims, or gather reliable citations for academic and technical writing.

This skill provides a systematic framework for evaluating the rigor of scientific claims, research methodology, and evidence quality. It guides users through assessing experimental design validity, identifying various types of biases (e.g., selection, cognitive), and applying established grading frameworks like GRADE or Cochrane Risk of Bias. Ideal for systematic reviews, peer review, and critical analysis of scientific literature.

This powerful skill audits the integrity and reliability of AI experimental results. It employs a cross-model review process to detect common forms of methodological fraud, including fake ground truth generation, suspicious score normalization, and insufficient scope reporting. Use this tool whenever you need an adversarial check on experimental claims to ensure the reported outcomes are robust and trustworthy.

Performs rigorous, zero-context verification of all quantitative claims, comparisons, and scope statements reported in a research paper against the raw evidence files. It is designed to detect confirmation bias, such as rounding errors, data cherry-picking, or misreported deltas, ensuring strict fidelity between published claims and underlying raw results. Ideal for pre-submission quality control.

This skill provides a single CLI interface to query numerous web search engines, including Brave, DuckDuckGo, Bing, Baidu, Sogou, and 360. It runs parallel searches, aggregates the results from various sources, and normalizes them into a uniform JSON format. Ideal for deep research, fact-checking claims across multiple platforms, or comparing global vs. regional search results.