agent-folder-init_skill

Overview

This skill initializes a comprehensive .agents/ folder structure tailored for AI-first development projects. I provide scaffolding for session tracking, task management, coding standards, ADRs, SOPs, security checklists, and agent-specific config folders for Claude, Codex, and Cursor. Use it to create a repeatable documentation and agent-config foundation that mirrors production patterns.

How this skill works

The scaffold script generates a documented .agents/ hierarchy with SYSTEM, TASKS, SESSIONS, SOP, EXAMPLES, and FEEDBACK areas plus daily session templates and task inbox files. It also creates agent config folders (.claude, .codex, .cursor) with commands, agents, and skills stubs copied from your library so projects receive up-to-date command sets while inheriting global rules. After generation, I expect you to customize PRD, ENTITIES, RULES, and SOP content to match the project.

When to use it

• Starting a new project that will rely on AI coding assistants and documented workflows
• Migrating an existing codebase to structured AI-first documentation and session tracking
• Onboarding teams that need consistent agent commands and session protocols
• Establishing coding standards, ADRs, and security checklists for projects using agents
• Creating reproducible templates for daily AI sessions and task management

Best practices

• Run the scaffold from your project root and then immediately fill SYSTEM/PRD.md and SYSTEM/ENTITIES.md
• Keep one session file per day (YYYY-MM-DD.md) and append multiple sessions inside the same file when needed
• Place project-specific command overrides in .claude/commands/ and put reusable rules in your global rules directory
• Use ALL-CAPS directories for top-level areas and ALL-CAPS for critical files to keep navigation consistent
• Integrate with linter, formatter, and CI skills after scaffolding to enforce standards automatically

Example use cases

• Bootstrapping a greenfield web app where engineers will collaborate with Claude and Codex agents
• Converting an ad-hoc project into a documented, auditable AI-driven workflow with SOPs and ADRs
• Creating a daily session log practice for a remote engineering team to capture agent interactions
• Standardizing command and agent behavior across multiple projects by copying commands from a central library
• Preparing a repo for audits by adding SECURITY-CHECKLIST.md and CRITICAL-NEVER-DO.md