Skill UI

Automates CI/CD pipeline setup, from linting and type checks to integration, E2E, security, and deployment stages, plus feedback loops for AI agents and preview deployments.

This skill provides comprehensive interaction with the DeerFlow AI agent platform via its robust HTTP API. It is designed for managing complex, multi-step AI workflows, including deep academic research, code execution, web browsing, and file analysis. Users can initiate and continue conversation threads, manage agent memory, upload documents (PDF, etc.), and delegate complex research tasks to the autonomous agent system.

Guides AI agents through analyzing working code, spotting overly complex structures, then refactoring for clarity while preserving exact behavior and project conventions, especially after reviews or high-pressure edits.

A comprehensive skill for generating non-generic, professional frontend interfaces, perfect for landing pages, portfolios, and product redesigns. This agent enforces a strict design methodology: reading the user's brief (audience, vibe, constraints) first, configuring global design dials (variance, motion, density), and recommending real, industry-standard design systems (like Material or Fluent UI) to ensure the output is always polished and contextually appropriate.

A robust CLI tool for connecting to Lark/Feishu real-time events. It allows streaming, subscribing, and consuming various events (e.g., IM messages, reactions, meeting updates) as NDJSON output. Ideal for developing Lark bots, building real-time message processors, or implementing long-running streaming webhook handlers, especially in AI agent subprocesses.

This guide details best practices for using `lark-cli` to interact with Feishu resources. It covers initial configuration, switching between user and bot identities, and crucial authentication flows for AI agents. It provides detailed instructions on handling permission errors, implementing secure split-flow authentication, and managing high-risk write operations to ensure data integrity and user consent.

This skill allows agents to ingest external resources (local files, URLs, repositories) using `ov add-resource`, and to create long-lasting searchable memories from conversations using `ov add-memory`. It is essential for building a comprehensive, persistent knowledge base for the OpenViking context database, ensuring that learned insights and external data remain available across sessions and interactions. Use when external knowledge injection or context persistence is required.

This mode simulates a P9 Tech Lead role, specializing in strategic project management and complex task orchestration. It does not write code itself, but focuses on generating detailed, structured Task Prompts (six elements) and managing the entire workflow delivery from multiple specialized agents (P8 teams). Ideal for coordinating large, complex projects requiring strategic decomposition and coordination of three or more parallel agents.

Customer Support Agent persona centralizes ticket tracking, email-to-task conversion, status logging, urgency escalation, and follow-up scheduling using Gmail, Sheets, Chat, and Calendar utilities.

This guide details how to systematically decompose large specifications or features into small, manageable, and verifiable tasks. It outlines best practices for vertical slicing, mapping component dependencies, and defining clear acceptance criteria. Use this process to ensure complex development work is broken down into reliably executable steps, suitable for agents or teams.

This meta-skill set provides a structured framework covering the entire software development lifecycle (SDLC). It guides developers through every phase, from initial concept refinement and requirement specification to incremental implementation, testing, code review, and deployment. It enforces critical professional behaviors like surfacing assumptions, managing ambiguity, and maintaining scope discipline to ensure high-quality, robust engineering outcomes.

This framework provides a rigorous, multi-dimensional guide for conducting code reviews. It evaluates five critical axes: correctness, readability, architecture, security, and performance. Use this process before merging any code change, especially when reviewing PRs, refactoring, or evaluating code produced by AI agents, to ensure continuous improvement and maintain high code health.