开发技能

Navigate the TRL library codebase with a guide to its SFT, DPO, GRPO, and KTO trainers, shared utilities, configuration, and model wrappers. Use it to understand expected contracts and data flow before inspecting or changing trainer code.

中文简介正在准备中。当前显示英语简介。

• Understanding the TRL trainer hierarchy
• Tracing data through a trainer workflow
• Checking utility function contracts before editing

Work with hex maps that use odd-r offset coordinates, where odd rows shift right. The skill explains how to find six neighbors, calculate distance between tiles, check adjacency, and list tiles within a chosen radius.

中文简介正在准备中。当前显示英语简介。

• Finding neighboring tiles on a hex map
• Measuring distance between two hex coordinates
• Checking whether tiles are adjacent

Use this guide to implement a quadrotor’s inner attitude loop. It explains how to turn desired horizontal acceleration into roll and pitch targets, then use PID feedback to calculate control moments, with guidance on integral state and gain adjustments.

中文简介正在准备中。当前显示英语简介。

• Implementing a quadrotor attitude controller
• Converting desired acceleration into roll and pitch targets
• Adding PID feedback for roll, pitch, and yaw

Implement a quadrotor’s outer position-control loop and plan trajectories from takeoff, hover, fly, and land waypoints. It covers PID feedback for desired acceleration and thrust, plus a sampled state matrix with position, velocity, orientation, angular velocity, and acceleration.

中文简介正在准备中。当前显示英语简介。

• Building a quadrotor position controller
• Generating trajectories from flight waypoints
• Tuning position PID gains for altitude response

Turn supported natural language commands for a drone simulator into timed waypoints and segment modes. The parser handles takeoff, hovering, movement between coordinates, and landing, returning arrays that a trajectory planner can use.

中文简介正在准备中。当前显示英语简介。

• Converting a takeoff instruction into timed waypoints
• Parsing hover duration and altitude commands
• Turning coordinate-based flight directions into segments

Decode signed numeric fields in mainframe data when the final character encodes both the sign and last digit. Use the lookup table and decoding steps to recover positive and negative values, including zero, and avoid treating debit amounts as positive.

中文简介正在准备中。当前显示英语简介。

• Reading signed amounts in converted mainframe files
• Decoding trailing letters or braces in numeric fields
• Checking positive and negative totals in fixed-length records

Use this skill to build and check mainframe-style batch programs with GnuCOBOL. It covers compile flags, sequential and fixed-length files, PIC fields, signed amounts, rounding, and a debug workflow for checking posting calculations.

中文简介正在准备中。当前显示英语简介。

• Compiling a free-format GnuCOBOL program
• Handling fixed-length sequential records
• Checking signed debit and credit calculations

Helps implement discrete PID feedback loops that calculate control output from current, accumulated, and changing error. It covers output limits, integral windup prevention, and practical gain tuning for systems such as vehicle speed regulation or other feedback controls.

中文简介正在准备中。当前显示英语简介。

• Implementing a PID feedback loop
• Tuning gains to reduce steady-state error or overshoot
• Managing throttle or braking output within limits

Read .xlsx workbooks with openpyxl or pandas, select the sheets and header rows you need, and convert rows into Python dictionaries. It also covers empty and merged cells and comma-separated values stored in a single cell.

中文简介正在准备中。当前显示英语简介。

• Extracting records from a multi-sheet workbook
• Turning spreadsheet rows into Python dictionaries
• Handling blank, merged, or list-valued cells

Use GitHub's official `gh` command-line tool to work with repositories, issues, pull requests, and workflow runs. It supports common tasks such as viewing diffs, checking CI status, creating or reviewing pull requests, and making authenticated API requests.

中文简介正在准备中。当前显示英语简介。

• Managing GitHub issues from the command line
• Viewing and reviewing pull requests
• Checking workflow runs and CI status

Compare feasible placements for jobs or tasks by estimating how much useful capacity each choice leaves behind. It helps with bin packing, GPU sharing, accelerator placement, and multi-resource scheduling, while weighing fragmentation alongside other priorities.

中文简介正在准备中。当前显示英语简介。

• Placing GPU jobs while preserving full slots
• Packing small tasks into bins efficiently
• Scheduling jobs with several resource needs

Use this reference to read COBOL COMP-3 packed decimal fields in input records, declare the right byte layout in an FD, and carry decoded values into accumulators. It explains sign nibbles, common layout mistakes, and checks for opening balances.

中文简介正在准备中。当前显示英语简介。

• Reading a packed balance from a COBOL input record
• Checking an FD declaration against the record layout
• Tracing an opening balance into an accumulator

Read and write YAML configuration files with safe parsing and clear error handling. The guidance includes readable output, preserving key order, handling missing or invalid files, and merging loaded settings over defaults, including vehicle parameters.

中文简介正在准备中。当前显示英语简介。

• Loading settings safely from a YAML file
• Writing readable YAML while keeping key order
• Using defaults when a config file is missing

Simulate vehicle motion over time by updating speed, position, and the gap to a vehicle ahead. Calculate a time-based safe following distance and estimate time to collision, then use those conditions to choose cruise, follow, or emergency modes.

中文简介正在准备中。当前显示英语简介。

• Simulating speed and position updates over time
• Calculating safe gaps behind a lead vehicle
• Estimating time to collision from relative speed

Use this skill to turn an MCP server’s tool, resource, and prompt inventory into versioned capability declarations and initialization options. It can also document supported features, set feature flags, handle capability negotiation, and generate capability tests.

中文简介正在准备中。当前显示英语简介。

• Declaring the features an MCP server supports
• Creating server initialization options
• Documenting tools, resources, and prompts

Create interactive command-line input flows with Inquirer.js, including validation, conditional prompts, and multi-step wizards. Generate prompt code, TypeScript answer types, reusable validators, formatters, and usage documentation.

中文简介正在准备中。当前显示英语简介。

• Building a command-line setup wizard
• Collecting and validating CLI configuration
• Showing prompts based on earlier answers

Design URI schemes and templates for resources exposed by MCP servers. It helps define resource hierarchies, generate code to parse and build URIs, map content types, support resource discovery, and document the resulting resource API.

中文简介正在准备中。当前显示英语简介。

• Plan URI patterns for a new MCP resource provider
• Generate helpers to parse and build resource URIs
• Document resource hierarchy and API templates

Detect and normalize CRLF and LF line endings in files shared across operating systems. The skill can handle mixed endings and help set Git configuration and `.gitattributes` rules so scripts and other files use the intended format consistently.

中文简介正在准备中。当前显示英语简介。

• Normalize files shared between Windows and Unix-like systems
• Check and repair files with mixed line endings
• Set Git rules for script and text file endings

Build terminal user interfaces with generated `Ink` components, using React patterns for layout, input, and state. Create interactive lists, text inputs, loading indicators, and custom hooks for command-line applications.

中文简介正在准备中。当前显示英语简介。

• Building an interactive terminal interface with React
• Creating selection lists or text input for a CLI
• Adding loading and progress indicators to a terminal app

Generate validation logic for related command-line arguments. Enforce options that cannot be used together, require one choice when needed, check dependencies between arguments, and provide clear conflict messages, with TypeScript and Python patterns.

中文简介正在准备中。当前显示英语简介。

• Rejecting conflicting command-line options
• Requiring one option from a group
• Checking dependent arguments

Turn MCP tool parameters into documented schemas that AI agents can interpret and applications can validate. Generate JSON Schema, TypeScript types, or Zod validation, including nested objects, enums, constraints, defaults, and example inputs.

中文简介正在准备中。当前显示英语简介。

• Defining input parameters for an MCP tool
• Generating TypeScript types from tool parameters
• Adding Zod validation for tool inputs

Scaffold an MCP server using the official TypeScript SDK, including transport setup, tool and resource handlers, input validation, and error handling. Choose stdio, SSE, or WebSocket transport and provide the server name and desired capabilities to shape the generated project.

中文简介正在准备中。当前显示英语简介。

• Starting a new MCP server project
• Adding MCP capabilities to an existing project
• Scaffolding tool and resource handlers

Create consistent, readable help text for command-line applications. Organize usage, arguments, options, and examples into clear sections, apply styling and indentation, and produce output that can also support man-page-compatible documentation.

中文简介正在准备中。当前显示英语简介。

• Formatting help output for a CLI command
• Adding clear examples to command documentation
• Building a custom help renderer

Add command-line argument handling to shell scripts with generated `getopts` code. It can support short and long options, check required arguments and values, and provide help text and usage documentation.

中文简介正在准备中。当前显示英语简介。

• Adding command-line options to a shell script
• Checking option values and required arguments
• Generating help text and usage instructions