Skill UI

This comprehensive skill transforms raw game concepts into fully actionable, structured Unity development projects. It guides users through a five-phase pipeline, generating detailed Game Design Documents (GDDs), technical scene blueprints, core system architectures, and ready-to-use AI prompts for assets (3D models, textures, music, SFX). Ideal for concept validation and starting a complex game project from scratch.

Detects data exfiltration via URL path encoding and chained web_fetch navigation, covering fake trusted UI injection, indirect prompt injection, letter-level path exfiltration, and multi-hop navigation hijacking in agent memory/context scenarios.

This meta-skill provides a comprehensive framework for the entire lifecycle of AI agent skills. It supports creating, refactoring, rigorously evaluating (e.g., Skill OS 2.0 Gates), and securely packaging skills from various sources like workflows, prompts, and documents. It ensures standardized, team-ready, and governed deployment of advanced AI capabilities.

This skill optimizes LLM context usage through prompt caching, log pruning, and AST-based file inspection. It enforces ultra-dense output, minimizes filler text, and manages token budgets efficiently. Ideal for technical tasks requiring precise code edits, structured data handling, and reduced context window consumption during long sessions.

Compares prompt changes against a fixed evaluation set, recording model settings, checks, missing results, judge limitations, and case-level regression evidence for a reviewable decision.

Detects prompt injection hidden in legal documents (pleadings, skeletons, evidence) before AI reads them. Scans PDF, Word, HTML, email attachments for invisible text, Unicode smuggling, and AI-directed instructions. Fails closed with optional hook integration.

Access MCP (Model Context Protocol) servers and their tools, resources, and prompts through the mcpdo CLI. Offers commands for listing, connecting, inspecting, and scripting against MCP servers (stdio or HTTP), extending your capabilities with external MCP connections.

Analyze prompt structure to maximize LLM cache reuse, then measure input cost and latency. Covers stable prefixes, user/workspace boundaries, provider-specific cache counters and a cost comparison calculator for OpenAI, Claude and Gemini.

Composes or improves LLM prompts and agent instructions. Asks one round of clarifying questions, then returns only the improved prompt without executing the task. Ideal for crafting effective prompts for AI agents.

AI Elements is a component library built on shadcn/ui designed to help developers build AI-native applications faster. It provides pre-built components for conversations, messages, tool displays, and prompt inputs. Ideal for creating chatbots, AI assistants, or any AI-powered chat interface. Supports React and Next.js with easy CLI installation and full customization options using Tailwind CSS.

This skill detects and recovers text accidentally typed using the Taiwanese Bopomofo keyboard while the input method was stuck in English mode. It maps garbled keystrokes like su3cl3a8 back to Traditional Chinese characters using a decoder script and context analysis. It is designed for Taiwanese developers who forget to switch input methods when prompting AI, ensuring requests are understood without retyping. It excludes technical strings like code or URLs.

Statistical gate for agent, prompt, model and tool changes: repeated trials, per-case paired comparison against a baseline, trajectory grading, flake detection and blocking exit codes before shipping.