Agentic Coding

This GitHub repository provides comprehensive resources on advanced Claude Code practices directly aligned with your interests. It covers best practices for working with AI agents, including detailed guidance on using AGENTS.md files, Skills organization in .claude/skills/ directories, and PRD workflow patterns through the Research → Plan → Execute → Review → Ship architectural model. The repo features 83 actionable tips and tricks organized by topic including prompting, planning with specs, context management, session workflows, and advanced agentic patterns like subagents for context isolation and agent teams for parallel development. The content includes multiple development workflow methodologies (11 featured, including "Everything Claude Code" with 210k stars and "Superpowers" with 221k stars) that demonstrate command → agent → skill orchestration patterns, plus curated collections of 25+ skill repositories and agent libraries. Video and podcast resources highlight "From Vibe Coding to Agentic Engineering" (Andrej, May 2026) and "Full Walkthrough: Workflow for AI Coding" (Matt Pocock, April 2026) covering advanced techniques, while documented best practices address efficient prompt engineering, phase-wise gated planning with vertical slices, context rot management at 300-400k tokens, and permission handling through auto mode and sandboxing.

The user is interested in best practices for AI agents, efficient use of Claude Code with PRD documents and AGENTS.md files, and advanced agentic coding techniques. The website content directly addresses their intent. This article describes a system for fixing Claude's inconsistency across sessions by using an interview-based approach to build precise shared language. The author developed a /brief skill that conducts a six-question interview forcing one-sentence answers, then saves the resulting language to a context file that loads in each new session. This approach, inspired by Matt Pocock's /grill-me skill, addresses the problem of Claude drifting between sessions by replacing vague CLAUDE.md files with decision-tested language that gets progressively more refined. The system works across domains including code, research, and strategy, and the author emphasizes that the solution to AI inconsistency isn't more instructions but better language forged through interrogation, which creates a living context file that compounds precision over time.

The website content provides practical guidance on managing Claude Code's context window for larger projects, covering token optimization strategies like the 80/20 rule for task complexity, the /compact command for conversation compression, and techniques for maintaining persistent project knowledge through CLAUDE.md files. It explains context recovery methods when losses occur mid-session, monitoring approaches using the status bar and /context command, and handoff note strategies for continuity between sessions. The guide emphasizes breaking work into context-sized chunks with natural breakpoints, completing components before integration, and using checkpoint notes to preserve architectural decisions and patterns across session restarts.

Antigravity Awesome Skills is a large installable library of 1,525+ reusable SKILL.md playbooks designed for Claude Code, Cursor, Codex CLI, Gemini CLI, and other AI coding assistants. The repository provides structured operating instructions organized into specialized plugins, bundles, and workflows across development, testing, security, infrastructure, product, and marketing domains. Users can install the full library with `npx antigravity-awesome-skills` or choose focused specialized plugins for specific workflows like web apps, security, data analytics, DevOps, QA, OSS maintenance, and agentic systems. The project includes role-based bundles for outcomes like shipping a SaaS MVP or production hardening, machine-readable workflows for ordered execution, and a searchable hosted catalog. Version 12.2.1 is trusted by 40,000+ GitHub stargazers and integrates official skills from Anthropic, Vercel Labs, OpenAI, Supabase, Microsoft, Google Gemini, Apify, and over 100 community contributors.