Skill UI

Hivemind is an advanced orchestration layer designed to manage complex tasks using a parallel swarm of external, low-cost workers. It simulates a structured team workflow (scout, coder, tester) to decompose large problems, execute subtasks in isolated worktrees, synthesize aggregated results, and perform consensus reviews. This dramatically reduces token costs and enhances reliability for large-scale coding, research, and testing tasks.

A guided tool to manage and progress through a structured, multi-module technical learning curriculum. It tracks prerequisites, requires recording concrete evidence artifacts for completion, and implements spaced repetition reviews to ensure long-term knowledge retention in coding practices.

This skill systematically audits Playwright End-to-End (E2E) test files. It checks for common anti-patterns, adherence to best practices, and identifies coverage gaps across the entire test suite. It provides a detailed score (1-10) and generates actionable reports, including suggested code fixes, significantly improving test reliability and maintainability.

This skill analyzes code for language and framework-specific security best practices. It is designed to help users write secure-by-default code, passively detect major vulnerabilities, or generate comprehensive, prioritized vulnerability reports. Supports languages including Python, JavaScript/TypeScript, and Go, ensuring robust security standards throughout the development lifecycle.

Tinte is an agent-native infrastructure designed for generating, compiling, and managing consistent design systems from a single source of truth. It maintains a graph of 13 semantic OKLCH color tokens, compiling them into multiple formats including shadcn/ui presets, VS Code themes, and terminal configurations. Use it to ensure brand consistency, extract themes from images (via AI), and generate code previews with professional styling.

A comprehensive guide and set of rules for writing effective, warm, and actionable comments across various development platforms. It helps users structure feedback for GitHub PRs, issue trackers, or Slack messages, ensuring the tone is consistently thoughtful, direct, and focused on clear next steps, rather than simply listing observations.

This skill analyzes excessively large Pull Requests (PRs) that exceed established line budgets (e.g., 400 lines). It automatically splits these massive changes into smaller, focused, and sequential PRs using advanced chaining or stacking strategies. This process is crucial for protecting reviewer focus, ensuring reviews can be completed within a manageable time budget (e.g., ≤ 60 minutes), and maintaining high code quality during complex feature integration.

A rigorous, standalone tool designed for explicit, blind, and adversarial code review. It executes the code target through two parallel, independent judges (Judge A and Judge B) with strict rules and limited fix rounds. This process ensures maximum scrutiny, providing a definitive, highly corroborated verdict (APPROVED or ESCALATED) that replaces standard review methods.

This document defines the critical architectural contract for standardizing advisory transport across diverse AI code review runtimes (e.g., Claude Code, OpenCode, Codex). It mandates that all integrations adhere to a shared Go contract, governing how review results, evidence, and validation are processed. This is essential for ensuring system integrity, managing architectural transitions, and preventing the creation of isolated, proprietary review systems.

This skill outlines a highly rigorous, receipt-driven workflow for managing defects and implementing code corrections. It guides public collaboration by enforcing strict controls over change authority, ensuring comprehensive lineage tracking, and mandating evidence-based validation. Key steps include CodeGraph impact mapping, black-box testing, and adherence to strict change budgets, ensuring every deployed change is fully traceable and robust.

Turn a plain "implement X for me" request into a working artifact: build a running end-to-end spine first, then add one feature per rung, recording every under-determined decision in an assumption ledger before its dependent code, with cross-model review to catch the rest.

Design AI agent architectures through requirements discovery, or audit and diagnose architectural flaws in existing agents. Architecture only; excludes implementation and general code review. Deliver justified decisions, verification methods, PDF reports, and visualizations.