AI: agentic workflows
This course covers Claude Code, Anthropic's coding agent, on a Rails app. It explains the agent loop of gathering context, acting and verifying, and the permission modes that set how much it does without asking. Lessons cover project instructions in CLAUDE.md, plans and tests as the definition of done, and checkpoints. They also cover writing skills, the Superpowers plugin method, review loops with /code-review and /simplify, and guardrails built from permission rules, hooks, the sandbox and worktrees. A final lesson runs one coupon feature from idea to pull request.
What you'll learn
- ✓How Claude Code runs the agent loop and how permission modes control its autonomy
- ✓Writing CLAUDE.md, checkable prompts and plans, and using rewind and context commands
- ✓Writing skills with SKILL.md and descriptions that route requests to them
- ✓Running the Superpowers method: brainstorming, plans, subagent execution and discipline skills
- ✓Review loops with tests, RuboCop, Brakeman, /code-review and /simplify
- ✓Permission rules, hooks, the sandbox, worktrees and headless runs
Claude Code: an agent in your terminal
In the agents course you built a loop by hand. The model asked for a tool, your code ran it, the result went back into the conversation, and the loop continued until the model said it was done. That loop fits on one screen, and it is the whole idea behind every coding agent.
Read the full lesson free →Curriculum
7 lessonsClaude Code: an agent in your terminal
Introduces Claude Code's loop of gathering context, acting and verifying, its tools, installation and permission modes.
Driving the loop on a Rails app
Covers CLAUDE.md, prompts with a checkable definition of done, plan mode, rewind limits and keeping context clean.
Skills: procedures on demand
Shows how to write a SKILL.md, how skills load in layers, and how descriptions route requests to them.
Superpowers: a workflow made of skills
Explains the Superpowers plugin, its brainstorming, planning and execution skills, and its rules for tests, verification and debugging.
Review loops: /code-review and /simplify
Builds a review order from deterministic checks, /code-review for bugs and /simplify for cleanup, as an evaluator-optimizer loop.
Guardrails: permissions, hooks and worktrees
Covers allow, ask and deny rules, Bash pattern matching, hooks and exit codes, the sandbox, worktrees and headless runs.
Putting it together: a feature end to end
Runs a coupon code feature through every layer, from worktree to pull request, and sets a permission evaluator challenge.