Skill UI

This guide details the comprehensive workflow for distributing a single AI skill across various local agent skill directories (Codex, Claude, Pi, Hermes). It addresses canonical locations, symlink management, and specific synchronization commands to ensure that all agents detect the latest skill version while correctly navigating common development pitfalls.

This guide provides a comprehensive reference on Agent Skills, a standardized and powerful method for providing structured context, workflows, and tools to Large Language Models (LLMs). It details the anatomy of the SKILL.md file, the three-stage progressive disclosure loading process, and best practices for creating both capability (tool-based) and process (workflow-based) primitives. Use this resource when designing, reviewing, or debugging agent-driven applications to ensure reliable and context-efficient performance.

This skill helps developers create highly focused, folder-specific documentation (`CLAUDE.md` and `AGENTS.md`). It systematically analyzes all files in a target directory to extract project purpose, operational constraints, naming conventions, and critical decisions. This deep context ensures that future AI agents (like Claude or Codex) operate with localized knowledge, preventing them from relying on global, generalized instructions and improving their accuracy in that specific domain.

The /goal command converts standard agent prompts into persistent, self-correcting loops (plan → act → test → review → iterate). It is designed for long-running, complex engineering tasks like migrations, coverage lifts, or refactoring, ensuring progress is tracked against a verifiable stop condition and mandated documentation.

This guide explains how to register custom model slugs (variants) within the Pi Agent's configuration files. It is essential for ensuring that specialized OpenRouter variants or brand-new provider models, which are not part of the static bundled registry, resolve correctly and are available for accurate agent model invocation.

A comprehensive suite designed for AI agents to perform safe, deep, and multi-format web searches. It allows agents to retrieve current web information, analyze PDFs, extract YouTube transcripts, and access GitHub code files. Instructions include guidelines for varying search depths and using specialized tools for code context retrieval.

This skill executes the DeepSWE benchmark, a comprehensive coding-agent evaluation suite. It uses OpenRouter and mini-swe-agent to score various Large Language Models (LLMs) on 113 tasks, providing reproducible results for model comparison, development assessment, and research benchmarking. Use this when a standardized, deep evaluation of LLM coding capabilities is required.

This skill acts as a critical reality-check layer before code generation. It ensures that every function call, import, method, and variable the agent uses is provably available in the user's specific installed environment and version. It prevents runtime bugs caused by hallucinated or improperly assumed APIs, making the generated code robust and reliable.

This skill utilizes the Taisly Agent Kit to manage the entire workflow for publishing approved short-form videos across major platforms, including TikTok, Instagram Reels, YouTube Shorts, X, and Facebook. It facilitates agent-assisted publishing by ensuring coordination of captions, metadata, and campaign details. A critical feature is the mandatory explicit confirmation gate, which prevents any state-changing publishing action without explicit user approval, prioritizing safety and compliance.

This skill guides the secure management and operation of Virtual Private Servers (VPS) and deployed server-side agents. Use it when SSH access, system restarts, log inspection, or complex workflow automation (like n8n or custom agents) is required on authorized remote hosts. It emphasizes following strict operational protocols and obtaining explicit user permission for sensitive actions.

WGM is a robust, agentic methodology designed to transform vague or rough requirements into working, validated software. It enforces a disciplined workflow across six stages: deep alignment (Grill), structured planning, deterministic iterative building (Loop), mandatory holdout-scenario judging, and final documentation audits. It ensures that the resulting code converges against rigorously tested acceptance criteria rather than relying on single-pass generation.

Automates the creation of Architecture Decision Records (ADRs). This tool manages the entire lifecycle of an ADR: determining the next sequential ID, generating a standardized Markdown file, storing the record in the AgentDB graph, and automatically linking it to related decisions. Essential for documenting significant architectural choices, API designs, and data model changes to maintain system knowledge coherence.