Coding & Development

Show native UI from scripts and agents — dialogs, forms, visualizations, floating widgets, cursor companions. Supports macOS, Linux, and Windows. Use when you need to display HTML to the user, collect input, show a chart, render markdown, or create any visual interaction without a browser.

Find what in a git repo needs refactoring, and what doesn't, the way developers used to feel it while coding by hand. It measures where changes keep landing (a heatmap of git history), what makes those files hard to change (code smells, each with its playbook of named refactorings) and what the next change would cost, then ranks refactors on a checklist worked through one item at a time. Use it whenever the user asks for a refactor pass; asks about tech debt, code health, hot spots or cleaning up; asks for refactor opportunities, or code that's cleaner, more maintainable, scalable or SOLID; asks what the worst part of the codebase is; says "we've made a lot of changes, anything to…

Review local Git changes with an evidence ledger that exposes unreviewed files, unsupported claims, and stale conclusions. Use for a pre-commit review or an explicit review blind-spot check. Not for general implementation or claims about benchmark accuracy.

Improve a GitHub repository's public presentation and discovery through README structure, banners, real showcases, accurate badges, installation guides, documentation navigation, About text, topics, and social preview assets. Use for repository makeovers, README redesigns, launch presentation, or repo visibility audits. Work across apps, tools, libraries, and AI agent skills, with extra guidance for skill distribution. Do not use for application code cleanup, refactoring, or growth guarantees.

Finds the bugs a test suite misses by using an app the way a person does, then turns each confirmed bug into a failing end-to-end test and a written bug entry. It hunts one area at a time with a hostile-user checklist (Undo every step, keyboard only, narrow window, dark mode, offline, big and broken files, format round trips, two tabs), runs automatic checks on every screenshot (contrast, clipped or overlapping text, errors), adds rules that every step must keep (Undo is exact, save and reopen gives the same file, web matches desktop), and keeps a rotation so every area gets hunted again. Use it whenever someone wants to find bugs, hunt for bugs, do exploratory or real-world testing, test...

仅在用户明确要求 Ant Design / antd 设计语言、点名 $antd-admin-template，或已明确授权当前项目设计必须使用本技能时使用。提供 React、TypeScript 与 Ant Design 的设计系统、组件复用、页面配方和模板；普通前端、后台、文档、看板任务以及已有 antd 依赖都不构成调用授权。

Choose the right Claude model (Haiku 5.5, Sonnet 5.5, Opus 5.5, Fable 5.2) and effort level for a task, subagent, or API integration. Use when the user asks which model to use, wants to cut cost or latency, is configuring subagent models, mixing models in one system, or migrating a request from one Claude model to another.

Use when asked to audit a repo's readiness for AI coding agents, or to review its AGENTS.md/CLAUDE.md, in-repo skills, lint guardrails, or commit/push gates. Read-only audit; outputs a report.

Build, review, debug and verify KOReader plugins (*.koplugin, Lua) for e-readers (Kobo, Kindle, PocketBook, Android) and the desktop emulator. Use whenever the user mentions KOReader, koplugin, koreader plugin, _meta.lua, WidgetContainer, registerToMainMenu, addToMainMenu, Dispatcher actions, UIManager, InfoMessage/InputDialog/Menu widgets, ./kodev, or wants to add a feature/menu item/gesture action to KOReader — even if they only say "e-reader plugin" or "make something for my Kobo/Kindle". Includes a scaffolder and a static validator for plugin structure, syntax and luacheck.

Explain code in plain language while building with an AI agent, so a person without a coding background (a designer, PM, founder or anyone else building with AI) gets a CS-level, big-picture understanding of code and architecture without needing to write code. Labels every piece by category (built-in, library, framework, service, custom code and so on), importance (core vs trivia) and complexity, and teaches the software landscape developers take for granted. Use whenever the user wants to learn, be taught, understand code "so I actually get it", start a "learning session", follow their learning path or learning doc, asks what something in the code is or how important it is, asks what cat...

Improve the Agent Experience (AX) of a website, API, SaaS app, or store, so that AI agents can reach it, discover it, read it, call it, get access, and recover from errors with no human present. Use when the user wants to make a product agent-ready or agent-friendly, or wants to add llms.txt, Markdown for agents, an API catalog, a better OpenAPI file, an MCP server, WebMCP, OAuth for agents, self-serve API keys, structured errors, robots.txt rules for AI bots, or agentic commerce. Also use to audit a URL, to get an AX Score from the bundled scanner, to review the design of an API or a feature for agents, or to add a CI gate on the score. Not for SEO ranking, not for citations in AI answer...

Interactively scaffold a production-minded FastAPI backend using Python 3.12, uv, and a chosen architecture (modular layered, clean, hexagonal, minimal, microservices), with optional PostgreSQL, Redis, Qdrant, JWT auth, Docker Compose, and AI modules (LangChain, LangGraph, RAG, MCP). Use when the user wants to start a new FastAPI project, scaffold a backend, or invokes this skill by name.

Build a 3D simulated environment (robotics / RL / autonomy scenes - self-driving streets, robot arms, drones, legged robots, warehouses, homes, farms, marine, space, sports, multi-agent) in MuJoCo or NVIDIA Isaac Sim from a description. Chooses the backend, confirms the spec with the user, sources 3D models from Poly Pizza (polypizza MCP) with procedural fallbacks, builds a tested scene + Gym-style env, and smoke-tests rendering, visuals and physics. Use whenever the user asks to create, generate, build or set up a simulation environment, sim scene, virtual world, RL/robot environment or digital twin.

Orchestrates the mcp-cura and mcp-moonraker MCP servers so an agent can take a 3D model from a mesh file to a supervised Klipper/Moonraker print: mesh inspection and repair, CuraEngine profile resolution and slicing, G-code safety verification, transfer to the printer, print start, live monitoring, pause/resume/cancel and job-queue management — always behind the servers' risk and approval gates. Use this skill whenever the user mentions slicing, 3D printing, STL/3MF slicing, CuraEngine, Klipper or Moonraker — even if they never name these MCP servers — and whenever the user asks to prepare, verify, send or supervise a print job. Do not use it for general CAD authoring or for editing Klipp...

Correct Godot 4.3+ TileMapLayer code - set_cell(coords, source_id, atlas_coords), reading cells and custom data, erasing, coordinate conversion, terrains, tile collision through TileSet physics layers, finding the tile a body hit, and migrating from the deprecated TileMap node and its layer-index API. Use when placing, reading or erasing tiles from GDScript, building a TileSet in code, generating levels procedurally, handling tile collisions, or porting TileMap code from Godot 3 or Godot 4.0 to

Fixes Godot 3 APIs and GDScript syntax that AI assistants write into Godot 4 projects, such as KinematicBody2D, move_and_slide(velocity), export, onready, setget, yield, string connect(), instance(), rand_range, the Tween node, File, change_scene, Spatial, rect_* properties, Pool*Array, OS time functions and the old TileMap API. Use when writing, reviewing or porting GDScript for Godot 4.x, when a script fails with "not declared", "not found in base", "was removed in Godot 4" or "is not present

Builds Godot 4 Control UIs that resize instead of clipping or overlapping - containers versus anchors, anchor presets from code, size flags, custom_minimum_size, MarginContainer margins, Label wrapping, mouse_filter, CanvasLayer HUDs, Godot 4 theme override names (add_theme_color_override and friends) and the display/window/stretch settings for multiple resolutions. Use when creating or fixing menus, HUDs, dialogs and inventory screens, when UI elements are cut off, zero-sized, overlapping, igno

Analyze repository architecture and stack to discover, configure, and orchestrate specialized agent roles (.agents/team/), audit existing roles for activity/pruning, and enforce human-in-the-loop coordination rules.

Godot 4 signal patterns and their failure modes - declaring typed signals, emit(), connect() and disconnect() with Callables, bind() and unbind(), lambdas, CONNECT_ONE_SHOT and CONNECT_DEFERRED, await on signals and timers, editor-made connections stored in .tscn files, and the call down, signal up rule. Use when wiring UI buttons, timers, Area2D or body events and custom game events, when replacing Godot 3 connect("signal", self, "method") or yield code, or when debugging "already connected", "

Structures Godot 4 projects so scenes stay reusable and references do not break - when to make a scene, a script or a Resource, instancing with @export PackedScene, preload and load, @onready and %UniqueName node access, _ready order, groups, autoloads, class_name, and resource UIDs with the .uid files added in Godot 4.4, including how to move or rename files safely. Use when creating new scenes or scripts, deciding how nodes find each other, adding global state or an event bus, moving or renami

Inspect repository meta-knowledge integrity, verify 100% boundary coverage, detect architectural drift, and opportunistically backfill missing facts without re-initialization.

Idiomatic Golang design patterns — functional options, constructors, error flow and cascading, resource management and lifecycle, graceful shutdown, resilience, architecture, dependency injection, data handling, streaming, and more. Apply when explicitly choosing between architectural patterns, implementing functional options, designing constructor APIs, setting up graceful shutdown, applying resilience patterns, or asking which idiomatic Go pattern fits a specific problem.

Common design patterns with Java examples (Factory, Builder, Strategy, Observer, Decorator, etc.). Use when user asks "implement pattern", "use factory", "strategy pattern", or when designing extensible components.

Scaffold and initialize the meta-project agent environment in any repository: checks for existing setup to safely upgrade without data loss, or runs a 2-pass scan to create AGENTS.md, .agents/project_knowledge/, and install skills.