Agentic Coding

The website content presents a comprehensive Claude Code plugin collection directly aligned with your workflow needs. The blueprint-plugin offers 30 skills specifically designed around PRD/PRP workflow with version tracking, which matches your interest in using PRD documents efficiently. The collection includes 14 agents across multiple categories—notably agents-plugin provides task-focused agents for test, review, debug, and docs workflows, while agent-patterns-plugin covers multi-agent coordination and orchestration patterns. For debugging and code quality, the code-quality-plugin provides 13 skills for code review, refactoring, and debugging methodology, and evaluate-plugin includes 3 agents for skill evaluation and benchmarking. The tiered setup approach with a Plugin Map and recommended install order provides structured guidance for advanced agentic coding workflows, and the /health:check and /health:audit skills offer diagnostic tools to optimize your Claude Code configuration.

The website content discusses skill-indexer, an open-source CLI tool for distributing AI agent skills across multiple coding assistants including Cursor and Claude Code. The tool addresses a key infrastructure challenge: how to share SKILL.md files across different AI coding platforms without each tool creating its own marketplace. The author outlines several unsettled design questions including whether npm is the right distribution substrate versus a dedicated registry, whether to use convention-based or declarative skill discovery, and how strictly to validate skill files to balance ecosystem quality with experimental skill inclusion. This directly addresses best practices for managing AI agent skills and workflows in tools like Cursor and Claude Code, which aligns with the user's interest in efficient agentic coding practices.

# AI Agents - Agentic Coding This GitHub repository curates a collection of Agent Skills—modular, standardized instruction packages that teach AI agents how to perform specific tasks in software engineering. Agent Skills are structured as SKILL.md files defining when to use them, how to execute tasks, and what best practices apply. They follow a progressive disclosure model where lightweight metadata loads initially and full instructions load only when relevant, keeping agents fast and focused while maintaining execution depth. The repository organizes skills across multiple engineering domains including Frontend Development, Backend Development, DevOps, Testing & QA, Security, Mobile Development, Blockchain, Infrastructure, Data Science, AI/ML, Observability, Performance, Tooling, and Code Review. Notable skills relevant to advanced agentic coding include command-development for Claude Code slash commands, implementing-agent-modes for PostHog AI agents, pair-programming for AI-assisted collaborative workflows, agentic-eval for evaluating and improving agent outputs, and loki-mode which orchestrates 100+ specialized agents from PRD to deployed product. The repository distinguishes Agent Skills from saved prompts, fine-tuning, and MCP tools—skills teach how to work while MCP provides access to external tools, making them complementary approaches for building sophisticated agent workflows.