10. Commands, Skills, and Capstone

The session that ran itself

With the agentic SDLC workflow I have built last year, I experimented with handing off tasks to my agents before going to bed. By the time I woke up, the tasks were completed, new plans were created as stored md files in my repo for me to review and GitHub PRs were opened for the refined issues with all the quality gates applied and production-grade code produced.

claude -p "/pick @specs/refined-issue.md"

claude -p "/plan adding a new notifications feature where users get notified with email when they finalized the course"

Those night sessions were built on top of everything this course teaches:

• A careful multi-step software development workflow and agent orchestration from Chapter 8 was triggered.
• The permission system from Chapter 4 allowed the sessions go smoothly without any interruptions during the sessions.
• The project memory and guardrails from Chapter 5 gave Claude project context and first line of feedback by catching type errors on every edit.
• The feedback loop from Chapter 6 gave Claude eyes and ears to feel and verify the outcome of the code it produced.
• The TDD workflow from Chapter 7 led the agents to focus on specs and covering existing and new functionalities.
• The context management from Chapter 9 kept the session lean and cost-efficient.

The commands and skills in this chapter are the layer that turns all of those pieces into repeatable, one-command automations.

This is also the final chapter of the Essentials course, and it closes with a capstone: you will build a complete feature end-to-end using the full toolkit.

Additionally, you will get access to all the reference prompts, configurations, subagents, skills and scripts we covered throughout the course.