数据与 AI技能

Build applications that exchange live audio, video, and text with the Gemini Live API. Covers voice conversations, streaming transcription and translation, session recovery, authentication, and migration from older Live models in Python or JavaScript/TypeScript.

中文简介正在准备中。当前显示英语简介。

• Build an AI app for live voice conversations
• Add real-time transcription or speech translation
• Send camera or screen frames alongside audio

Implement Gemini API applications in Python or TypeScript for text, image, video, and speech generation. Covers ongoing conversations, external tool calls, structured responses, background research, managed agents, and migration from the older generateContent API.

中文简介正在准备中。当前显示英语简介。

• Build a chatbot that retains conversation context
• Add image, video, or speech generation to an app
• Connect external tools and return structured data

Interpret the saved results of a completed AI council review. Count verdicts, check whether enough participants responded and whether models are independent, surface substantive dissent, and explain what action the outcome supports.

中文简介正在准备中。当前显示英语简介。

• Explain a completed council review
• Check vote counts and quorum
• Assess model independence and dissent

Compare estimated AI model costs for a task using its complexity and expected text volume. Get a suitable low-cost recommendation, a cheaper but riskier alternative, and a premium option, with explicit assumptions behind the estimates.

中文简介正在准备中。当前显示英语简介。

• Choosing a model for a bulk code refactor
• Comparing model costs for a security review
• Estimating costs for long-document analysis

Assess whether a multi-agent workflow justifies its coordination costs. Examine handoffs, lost context, and distinct contributions, then compare the setup with one capable agent to recommend keeping, simplifying, or restructuring it.

中文简介正在准备中。当前显示英语简介。

• Reviewing a workflow before adding another agent
• Finding costly handoffs in a slow workflow
• Checking whether agent agreement adds useful evidence

Turn a rough request into a structured AI prompt with a clear goal, output format, and constraints. For complex work, add smaller tasks, separate creator and reviewer roles, and explicit checks for accuracy and uncertainty.

中文简介正在准备中。当前显示英语简介。

• Refining a vague AI request
• Designing prompts for API documentation
• Adding verification steps to code generation

Gather advice from multiple language models assigned different roles, then combine their proposals and critiques into a recommendation. Set budget and depth limits while retaining disagreements, minority views, and risks for decisions or implementation planning.

中文简介正在准备中。当前显示英语简介。

• Comparing options before a decision
• Reviewing code or strategy from several perspectives
• Preparing an implementation plan with risks

Add GitHub Copilot CLI as an optional research provider in Claude Octopus workflows using an existing Copilot subscription. The skill covers authentication, model selection, and skipping Copilot when unavailable; each prompt consumes a premium request.

中文简介正在准备中。当前显示英语简介。

• Adding Copilot to research with multiple AI providers
• Using Copilot for code explanations and exploration
• Configuring Copilot authentication for CI

Rewrite a prompt to address the same task more deeply. The skill identifies what an expert would examine, which original materials and methods they would use, and what independent evidence would confirm the result.

中文简介正在准备中。当前显示英语简介。

• Refining a request without expanding its scope
• Identifying original materials and expert methods
• Adding independent criteria for checking results

Design, debug, and tune OpenRouter API integrations. It checks model capabilities, reasoning settings, output limits, and fallback routes, then validates results against quality, cost, privacy, and performance requirements, including failed attempts.

中文简介正在准备中。当前显示英语简介。

• Design an OpenRouter API integration
• Investigate empty, truncated, or malformed responses
• Configure fallback models and provider restrictions

Revise wording in a prompt that may cause Fable to mistakenly refuse a legitimate request. It preserves the goal and structure, returns the complete edited prompt, and lists sentence changes without guaranteeing acceptance.

中文简介正在准备中。当前显示英语简介。

• Rewording legitimate requests mistakenly refused by Fable
• Reviewing wording in authorized security tasks
• Adjusting medical or lab-related prompt wording

Query the H2 database created by tika-eval to get counts and comparisons missing from standard reports. Includes connection guidance and SQL examples for comparing extracted text, tracking encoding changes, and joining related file records.

中文简介正在准备中。当前显示英语简介。

• Get counts missing from standard reports
• Compare extracted text before and after a change
• Investigate changes in detected encodings

Explore scientific data files to understand their structure, statistical patterns, and quality. References cover over 200 formats, with workflows for producing Markdown reports that summarize findings and recommend preprocessing, visualizations, and further analysis.

中文简介正在准备中。当前显示英语简介。

• Checking missing values and outliers in experimental tables
• Examining sequence lengths and quality scores in FASTQ files
• Inspecting microscopy image dimensions and metadata

Prepare research data figures for journal submission with consistent layouts, error bars showing uncertainty, and accessible colors. Adjust figure dimensions, typography, resolution, and export formats using matplotlib, seaborn, and plotly.

中文简介正在准备中。当前显示英语简介。

• Preparing figures for journal submission
• Combining multiple plots into one consistent figure
• Improving color and grayscale readability

Estimate study sample sizes and statistical power, the chance of detecting a specified effect. Calculate detectable effects for a fixed sample, compare assumptions with power curves, and document adjustments for dropout, multiple comparisons, or clustered observations.

中文简介正在准备中。当前显示英语简介。

• Estimate participant or sample counts before data collection
• Find the smallest effect detectable with a fixed sample
• Compare sample requirements across plausible effect sizes

Analyze research data with guidance on choosing statistical tests, checking their assumptions, and measuring effect sizes, meaning the magnitude of differences or relationships. It also supports sample size planning, Bayesian analysis, and APA-style reports.

中文简介正在准备中。当前显示英语简介。

• Comparing groups in experimental or survey data
• Choosing tests and checking their assumptions
• Planning the required study sample size

Build interactive maps with AntV L7, a library for displaying geographic data. The skill guides data loading, points, routes, areas, heatmaps, map interactions, animations, and performance tuning for large datasets.

中文简介正在准备中。当前显示英语简介。

• Display locations from GeoJSON or CSV
• Animate routes on a map
• Show geographic density with a heatmap

Get a second opinion from Google Gemini on Google Ads, SEO metadata, or code changes. Choose a pass/fail review, a challenge that probes weaknesses, or an open consultation on strategy and implementation.

中文简介正在准备中。当前显示英语简介。

• Getting a second opinion on Google Ads changes
• Reviewing SEO metadata updates
• Probing weaknesses in proposed code changes

Turn CSV, Excel, or JSON data into a complete HTML report with key metrics, multiple charts, written insights, and a data table. Embedded charts let readers open the report offline.

中文简介正在准备中。当前显示英语简介。

• Create a report from CSV or Excel data
• Build a dashboard of key metrics
• Summarize trends and unusual values

Turn numerical or relationship data into static chart images through the AntV API. The skill selects a suitable chart type, prepares the request, and returns an image link for comparisons, trends, proportions, or diagrams.

中文简介正在准备中。当前显示英语简介。

• Show monthly trends with a line chart
• Compare categories with a bar chart
• Visualize proportions with a pie chart

Try model or training changes in a copied Grug or Grugformer experiment, document the differences, and bring successful changes back into the base template while removing superseded experiment code and checking behavior.

中文简介正在准备中。当前显示英语简介。

• Try changes to attention or masking
• Experiment with optimizers or training loops
• Bring successful changes into the base template

Upload model weights to Hugging Face Hub with an OpenMDW 1.1 license and a required research warning. Check the published README and license file after uploading or updating to confirm that the required information is present.

中文简介正在准备中。当前显示英语简介。

• Publishing model weights to Hugging Face Hub
• Checking documentation after a model update
• Adding OpenMDW 1.1 license files and metadata

Run Marin canary checks for training health or daily experiments with small configuration changes. Monitor each job to completion, record evaluation results, and close daily runs with a version tag and log update.

中文简介正在准备中。当前显示英语简介。

• Check training health with an unchanged canary run
• Test one or two configuration changes in a daily run
• Monitor a ferry job through its terminal state

Profile JAX, Levanter, or Marin training to investigate slow startup, compilation, initialization, or throughput. Turn execution traces, recorded timing events, into structured reports, identify costly operations, and compare performance before and after a bounded change.

中文简介正在准备中。当前显示英语简介。

• Investigate slow training startup or compilation
• Identify costly operations and communication delays
• Compare training performance before and after a change