Agentic Coding

This repository documents advanced Claude Code practices directly aligned with your interests. It covers efficient work with AI agents using PRD documents, AGENTS.md files, and Skills organization. The content includes practical tips for prompt engineering, context management, and agentic workflows—such as breaking PRDs into vertical slices, using subagents for context management, and leveraging the Research → Plan → Execute → Review → Ship architectural pattern across development workflows. The repo features curated collections of skill libraries, agent definitions, and complete workflow methodologies (11 major workflows listed with star counts and implementation patterns). It includes 83 tips organized by category covering prompting strategies, planning with specs, session management, CLAUDE.md best practices, and advanced techniques like using multiple agents for parallel development and test-time compute. Notable resources include videos from Boris Cherny (Claude Code creator), Matt Pocock, and Andrej Karpathy on agentic engineering workflows, plus links to open-source workflow implementations like weather-orchestrator that demonstrate command → agent → skill patterns you can use as templates.

This session presents a validation-first engineering loop for AI coding assistants that moves beyond simple prompting. The approach emphasizes front-loading architecture context through documentation and project rules, structuring implementation as discrete tasks, and implementing multi-layered validation systems to catch issues before code review. Rather than relying on better prompts alone, the framework treats AI assistants as true engineers needing proper context and feedback mechanisms, allowing the system to evolve by addressing root causes when mistakes occur—enabling consistent delivery of production-ready code with any AI coding tool.