Use when the user wants to test their LLM/agent application for safety and security vulnerabilities — jailbreaks, prompt injection, PII extraction, harmful content generation, or evaluator gaming. Also use when the user mentions security testing, adversarial testing, red teaming, safety evaluation, ASR (Attack Success Rate), or "is my app safe to deploy." Outputs ASR paired with over-refusal rate and an audit document.
中文简介正在准备中。当前显示原文说明。
Build custom LLM evaluation pipelines using the OpenJudge framework. Covers selecting and configuring graders (LLM-based, function-based, agentic), running batch evaluations with GradingRunner, combining scores with aggregators, applying evaluation strategies (voting, average), auto-generating graders from data, and analyzing results (pairwise win rates, statistics, validation metrics). Use when the user wants to evaluate LLM outputs, compare multiple models, design scoring criteria, or build an automated evaluation system.
中文简介正在准备中。当前显示原文说明。
Build RL reward signals using the OpenJudge framework. Covers choosing between pointwise and pairwise reward strategies based on RL algorithm, task type, and cost; aggregating multi-dimensional pointwise scores into a scalar reward; pairwise tournament reward for GRPO on subjective tasks (net win rate across group rollouts); generating preference pairs for DPO/RLAIF; and normalizing scores for training stability. Use when building reward models, scoring rollouts for GRPO/REINFORCE, generating preference data for DPO, or doing Best-of-N selection.
中文简介正在准备中。当前显示原文说明。
Use when the user wants to build an evaluation system for an LLM/agent application but doesn't know where to start — they have traces, prompts, RAG pipelines, or nothing at all. Also use when the user mentions evaluation, eval, benchmarking, testing LLM quality, measuring agent performance, assessing RAG accuracy, or wants to compare prompts/models. This skill is the entry router: it asks diagnostic questions then recommends which sub-skill (local workflow) to use next.
中文简介正在准备中。当前显示原文说明。
Use when the user has a judge/grader and human-labeled data, and wants to measure how well the judge agrees with humans, detect systematic biases, determine whether automatic evaluation can replace human review, or build a human-reduction roadmap. Also use when the user mentions calibration, TPR/TNR, judge validation, inter-rater agreement, Cohen's kappa, bias detection, or "is my automatic evaluation trustworthy." Merges the calibrate and align functions into one skill.
中文简介正在准备中。当前显示原文说明。
Use when the user has nothing — no traces, no labels, no eval set — and needs to build a v0 evaluation from scratch. Also use when the user says "I need to start evaluating my app but don't know where to begin," "I want to set up eval for a new product," or has just identified failure modes and needs to turn them into principles. Outputs a v0 grader in 30 minutes using OpenJudge SimpleRubricsGenerator, plus a roadmap to reach calibrated evaluation.
中文简介正在准备中。当前显示原文说明。
Benchmark LLM reference recommendation capabilities by verifying every cited paper against Crossref, PubMed, arXiv, and DBLP. Measures hallucination rate, per-field accuracy (title/author/year/DOI), discipline breakdown, and year constraint compliance. Supports tool-augmented (ReAct + web search) mode. Use when the user asks to evaluate, benchmark, or compare models on academic reference hallucination, literature recommendation quality, or citation accuracy.
中文简介正在准备中。当前显示原文说明。
Detect whether an API endpoint is backed by genuine Claude (not a wrapper, proxy, or impersonator) using 9 weighted rule-based checks that mirror the claude-verify project. Also extracts injected system prompts from providers that override Claude's identity. Fully self-contained — copy the code below and run, no extra packages beyond httpx. Use when the user wants to verify a Claude API key or endpoint, check if a third-party Claude service is authentic, audit API providers for Claude authenticity, test multiple models in parallel, or discover what system prompt a provider has injected.
中文简介正在准备中。当前显示原文说明。
Benchmark LLM reference recommendation capabilities by verifying every cited paper against Crossref, PubMed, arXiv, and DBLP. Measures hallucination rate, per-field accuracy (title/author/year/DOI), discipline breakdown, and year constraint compliance. Supports tool-augmented (ReAct + web search) mode. Use when the user asks to evaluate, benchmark, or compare models on academic reference hallucination, literature recommendation quality, or citation accuracy.
中文简介正在准备中。当前显示原文说明。
Generate text, images, video, speech, and music via the MiniMax AI platform. Covers text generation (MiniMax-M3 model), image generation (image-01), video generation (Hailuo-2.3), speech synthesis (speech-2.8-hd, 300+ voices), music generation (music-2.6 with lyrics, cover, and instrumental), and web search. Use when the user needs to create AI-generated multimedia content, produce narrated audio from text, compose music, or search the web through MiniMax AI services.
中文简介正在准备中。当前显示原文说明。
Use when the user has a RAG (Retrieval-Augmented Generation) system and wants to evaluate its quality — separating retrieval issues from generation issues. Also use when the user mentions RAG evaluation, faithfulness checking, hallucination detection in RAG, retrieval quality, chunking optimization, or "is my RAG pipeline working." Outputs a diagnostic matrix that pinpoints whether problems are in retrieval or generation.
中文简介正在准备中。当前显示原文说明。
React and Next.js performance optimization guidelines from Vercel Engineering. This skill should be used when writing, reviewing, or refactoring React/Next.js code to ensure optimal performance patterns. Triggers on tasks involving React components, Next.js pages, data fetching, bundle optimization, or performance improvements.
中文简介正在准备中。当前显示原文说明。
Generates English pull request titles and bodies by comparing the current branch with main. Use when the user asks to draft PR content, summarize branch changes, or produce a PR title/body from git diff.
中文简介正在准备中。当前显示原文说明。
Guidelines for handling databases
中文简介正在准备中。当前显示原文说明。
Guidelines for handling text files
中文简介正在准备中。当前显示原文说明。
Guildlines for handling json files
中文简介正在准备中。当前显示原文说明。
Guidelines for handling image files
中文简介正在准备中。当前显示原文说明。
Guidelines for handling CSV/Excel files
中文简介正在准备中。当前显示原文说明。
Data Skill-Hub 技能仓库使用指南。在执行任何数据分析任务前，必须优先加载此技能以了解可用技能体系和调用规范。
中文简介正在准备中。当前显示原文说明。
Computes a numeric or categorical answer to a quantitative data-science question by cleaning and analyzing local data files (CSV, Excel, TSV, and scientific formats .npz/.fits/.h5) with pandas, numpy, and scipy. Use whenever a task ships its own dataset (in whatever local directory it provides) and asks you to derive a value via statistical, geospatial, temporal, ratio, percentage, growth-rate, correlation, signal-processing, orbital, or forecasting analysis. Spans domains including archeology (demographics, paleoclimate proxies, radiocarbon dating, historical conflicts); biomedical/bioinformatics (tumor histology, genes, proteins, peptides, genomic variants); environmental (water quality, bacterial/Enterococcus exceedance, rainfall correlations, environmental justice); legal/consumer (fraud, identity theft, consumer complaints by state/metro); wildfire (incidents, causes, suppression costs, acres burned, fatalities, geospatial intersections); and astronomy/heliophysics/space weather (satellite telemetry TLE/Swarm/SP3, indices OMNI/GOES, solar activity, atmospheric density, orbital mechanics, forecasting).
中文简介正在准备中。当前显示原文说明。
Builds end-to-end predictive modeling pipelines on tabular datasets—data exploration, preprocessing, baseline modeling, and mandatory optimization—producing valid saved predictions in the required output format. Use for any supervised machine learning or data modeling task, including classification and regression on tabular data, Kaggle-style competitions or submission-file challenges, training a model to predict a target from features, or requests to fit/tune/evaluate models with scikit-learn, XGBoost, Random Forest, Logistic Regression, and similar non-deep-learning methods. Runs fully automated end-to-end without asking the user for intermediate decisions.
中文简介正在准备中。当前显示原文说明。
对留存执行完整的分析流程，算清楚用户留不留、留多少、有没有变化，含口径定义、取数、留存率计算，及可选的维度拆分与对比/归因。触发条件：对话涉及留存率分析、留存表现评估、留存变化归因等任务时触发，如出现「留存表现情况如何」「访问留存和使用留存对比怎么样」「不同国家的访问留存是否存在差异」「分维度后留存情况」「次日/第七日留存率变化」等相似问题时触发。
中文简介正在准备中。当前显示原文说明。
对转化执行完整的分析流程，算清楚用户转不转、转化率多少、有没有变化，含口径定义、取数、转化率计算，及可选的异常分析、维度拆分与对比/归因。触发条件：对话涉及转化率分析、转化表现评估、转化变化归因等任务时触发，如出现「分析对话转化率」「转化率表现如何」「不同群体/地区/时间的转化率对比」「分维度后转化情况」「转化率是否异常」等相似问题时触发。
中文简介正在准备中。当前显示原文说明。
把用户按同一批/同一类分组，比较各群体后续表现。触发条件：对话涉及同期群分析、分群后表现对比、用户生命周期分群跟踪等任务时触发，如出现「同期群」「cohort」「按起始特征分群后表现如何」「不同客群后续转化/留存对比」「注册周/获客渠道分群跟踪」等相似问题时触发。
中文简介正在准备中。当前显示原文说明。