Antony Sure

Patterns for building agentic systems to enhance AI-powered product features.

Spawn parallel research subagents to investigate tech choices and libraries based on requirements.

Define system architecture based on approved PRD and research, producing comprehensive documentation.

Build Retrieval-Augmented Generation pipelines for effective document retrieval and context-based responses.

Establishes universal coding conventions for TypeScript, JavaScript, React, and Node.js.

UI audit and upgrade skill that identifies design issues and suggests structured fixes.

Minimalist Skill offers a clean, editorial-style UI aesthetic for content-heavy applications.

Execute plans with fresh subagents using TDD and a mandatory two-stage review process.

Create atomic execution plans for specific project phases with detailed task breakdowns.

Complete a feature branch with flexible merge options and automated cleanup.

Enforces complete output during implementation tasks to prevent lazy coding practices.

Trigger a manual code review of recent changes with a two-stage process.

Enforces RED-GREEN-REFACTOR TDD cycle during implementation. Write the failing test first, watch it fail, write minimal code, watch it pass, commit. Code written before tests must be deleted and restarted. Use when writing any implementation code within the Bayes Ship pipeline.

Enhance frontend design with a premium skill that ensures unique layouts, typography, and motion.

Implement a two-stage code review process to ensure spec compliance and code quality.

Extracts product requirements through structured Q&A and generates CONTEXT.md.

Generate a structured PRD from discovery and research outputs with automated review.

API design, query optimization, repository/service layer patterns for Node.js, Express, and Next.js backends. Use when implementing backend tasks. Source: everything-claude-code.

AI regression testing for model output drift, prompt sensitivity, and embedding consistency.

Manage package versions and dependencies for the LangChain ecosystem in Python and TypeScript.

Automate your build process with fully autonomous or semi-autonomous modes.

Automated security scanning against OWASP Top 10 vulnerabilities and secret detection.

Facilitates interactive requirements gathering through structured Q&A with users.

Resume from the last checkpoint after a session restart or compaction.

A premium UI aesthetic designed for fintech and professional tools, emphasizing trust and quality.

Evidence-based verification pattern. Never trust "it works" without running actual checks. Verify every acceptance criterion against observable behavior. Use before marking any task or phase as complete.

Automatically generates API documentation and changelogs from your codebase and commit history.

Creates atomic execution plans with file decomposition, task sizing, model routing, TDD blocks, and wave grouping.

Investigates a specific tech area based on project requirements. Writes findings to a research file.

Bayes Ship is an end-to-end product delivery system for Claude Code, streamlining the journey from idea to deployed prototype.

4-phase root cause debugging process. Gather evidence, form hypotheses, test hypotheses, confirm root cause. Use for any bug diagnosis within the Bayes Ship pipeline. Includes defense-in-depth and condition-based-waiting techniques.

Manage Git worktrees for isolated feature development and testing.

Initialize a new Bayes Ship product project quickly and efficiently.

Atomic Planning is a task decomposition methodology for Bayes Ship that streamlines project planning.

Stage 2 of two-stage review. Reviews code style, patterns, bugs, security, test quality. Only runs AFTER spec compliance passes.

Implement human-in-the-loop approval flows and custom middleware with LangChain.

Create complex, stateful AI workflows using LangGraph's StateGraph for multi-step processes.

Ensures implementation strictly adheres to specifications, identifying missing requirements and over-building.

Eval-driven development framework for AI features to ensure quality and reliability.

REST API endpoint design, request/response patterns, versioning, error handling, and authentication patterns. Use during architecture and implementation phases. Source: everything-claude-code.

Context rot prevention and subagent orchestration rules. Enforces clean context windows, file-based state, and delegation patterns. Auto-invoked during any Bayes Ship execution phase.

React and Next.js patterns for state management, component architecture, rendering, forms, and data fetching. Use when implementing frontend tasks. Source: everything-claude-code.

Build AI agents using LangChain's create_agent, tools, and middleware patterns.

Systematic bug diagnosis and fix using a structured 4-phase approach.

Automate production deployments with a comprehensive checklist and health verification.

Configure preferences for the Bayes Ship framework with customizable settings.

Display a comprehensive progress dashboard for project tracking.

Automate comprehensive QA verification for your projects with Ship Test.

Designs system architecture from PRD and research. Produces ARCHITECTURE.md, data model, API contracts, ADRs.

Automate deployments to Railway, Vercel, and other platforms with built-in checks.

Ensures PRD documents are complete, clear, and free of scope creep before user review.

Consolidates research files into a unified RESEARCH.md with clear recommendations and trade-offs.

Ensures comprehensive test coverage and performs AI-specific regression testing.

Systematic 4-phase root cause analysis. Reproduce, hypothesize, test, confirm. Includes defense-in-depth considerations.

Implement persistent memory and checkpointers for LangGraph agents to enhance conversation history and session durability.

Verification-before-completion agent. Runs actual checks against acceptance criteria. Evidence over claims.

Create isolated git worktrees for feature implementation with a verified test baseline.

WCAG accessibility patterns: ARIA labels, keyboard navigation, screen reader support, color contrast, focus management. Use during QA phase for any product with a UI. Source: everything-claude-code.

Writes code following TDD discipline. Receives a single task, writes failing test first, implements minimal code, reports status signal.