Agentic Coding

This comprehensive GitHub repository provides advanced guidance for using Claude Code and Cursor as an AI coding assistant. Key best practices include: starting with plan mode and using phase-wise gated planning with multiple tests per phase; breaking PRDs into vertical slices (tracer bullets) that cross all layers rather than horizontal phasing to enable earlier end-to-end feedback; using subagents for context management to keep main sessions focused; implementing skills in isolated subfolders for monorepos with progressive disclosure through references/, scripts/, and examples/ subdirectories; and managing context by rewinding failed attempts rather than leaving corrections in context, keeping under 30-40% context utilization for intelligence-sensitive work. The repository emphasizes the Research → Plan → Execute → Review → Ship architectural pattern across all major workflows, using commands for repeated inner-loop workflows, employing slash commands to avoid re-prompting, and leveraging cross-model approaches (e.g., planning in Claude, QA-review in another model) for higher-quality output. Advanced techniques include using PRD documents efficiently, maintaining lean CLAUDE.md files under 200 lines with domain-specific rules in .claude/rules/*.md, using agent teams with tmux and git worktrees for parallel development, and employing test-time compute with separate context windows where one agent identifies bugs and another finds them.

AGENTS.md is an agreed-upon standard for AI coding assistants like Cursor and GitHub Copilot that acts as a centralized manual defining your project's architecture, coding standards, common commands, key patterns, and documentation links. The file helps ensure AI suggestions align with your codebase, reducing manual corrections. For large projects, organize AGENTS.md hierarchically across directories—use a high-level root file with an overview and place detailed instructions in subfolder-specific AGENTS.md files, allowing the nearest file to take precedence. Best practices include being concise with important rules first, using imperative language with specific code examples, avoiding duplication with linter configs, and keeping instructions current. Note that Anthropic still uses CLAUDE.md instead of AGENTS.md but you can reference AGENTS.md within it. Most AI IDE tools can automatically generate and maintain this documentation.

Agent Skills is a security-hardened skill registry for professional AI coding agents like Claude Code, Cursor, Antigravity, and others. It offers verified, tested capabilities that extend agent functionality through packaged instructions and resources—similar to plugins for your AI assistant. The platform addresses marketplace vulnerabilities by providing 100% open-source skills with static analysis, human-curated prompts, and security scanning via Snyk before publishing. Featured skills include tlc-spec-driven for spec-driven development with persistent memory across sessions, aws-advisor for architecture and security guidance, and security-best-practices for vulnerability detection and secure-by-default fixes. Installation is managed through an interactive CLI (npx @tech-leads-club/agent-skills) that lets you select skills, choose target agents, and install globally or locally. An MCP server is also available to expose the skills catalog directly to AI agents with progressive disclosure for efficient browsing and loading.

This is a Git workflow skill package for AI agents like Claude Code and Cursor. It provides structured procedural knowledge for version control best practices including Git Flow and GitHub Flow branching strategies, Conventional Commits standards, pull request workflows, CI/CD integration, and advanced Git operations. The skill can be installed via marketplace, npm, composer, or git clone, and automatically triggers when users work on branching strategies, commit conventions, PR configurations, and merge conflict resolution. It includes practical commands like /pr-finish for automating pull request completion and comprehensive reference documentation on commit conventions, collaborative workflows, and CI/CD patterns—directly relevant for developers using Claude Code and Cursor who want structured Agent Skills to enhance their AI-assisted coding workflows.