Preporato
Build & submit taskBetaintermediate

Build a Claude Code Skill, Command, and Permission Set

Package a reusable workflow as a Claude Code Skill with frontmatter and progressive disclosure, add a custom slash command, and lock it down with a settings.json permission set (allowed tools and permission mode). Submit a single file for instant, rubric-based feedback.

2.5 hrs

Est. time

4

Outcomes

6

Rubric criteria

65%

Pass score

What you'll learn

Skills you'll have real reps in after shipping this.

Skills as packaged workflows
A Skill captures a reusable workflow with frontmatter and a body, so the whole team runs it the same way.
Progressive disclosure
Keep the skill body small and reference bundled detail on demand, so loading it stays cheap.
Custom commands
A slash command under .claude/commands/ is a fast, argument-aware entry point to a workflow.
Permissions and modes
Allowed and denied tools plus a permission mode set the guardrails for what the agent may do.

The scenario

Your team does the same multi-step workflow by hand in Claude Code every week (generate a migration, run checks, update the changelog). Everyone does it slightly differently, the instructions live in someone's head, and nobody has set guardrails on what the agent is allowed to touch while doing it.

You are going to package it. A Skill that captures the workflow with progressive disclosure so it stays cheap to load, a slash command that invokes it with arguments, and a permission set that says exactly which tools the agent may use and which mode it runs in.

Your role

You are a Claude solutions architect standardizing how the team extends Claude Code. Your deliverable is one file containing a production Skill, a custom command, and a permission configuration that are concrete and ready to drop into a repo.

Start the task to unlock the full brief

You'll get the step-by-step requirements, setup commands, the 6-criterion grading rubric, tips, and the ability to submit your solution for instant AI grading.

Free to start · submit when you're ready

What you'll build in this Claude Code skill task

This is a build-and-submit task, not a guided lab. You package a repeatable Claude Code workflow the way a team should: a Skill with frontmatter and progressive disclosure, a custom slash command that invokes it with arguments, and a permission set that fixes the guardrails. The deliverable is one file a team could drop into a repo and use unchanged.

The skills here are about making Claude Code consistent and safe to extend. You write a skill description specific enough to be discovered, keep the body small and reference detail on demand so it stays cheap to load, expose a fast command entry point, and lock down which tools the agent may use and which permission mode it runs in.

Grading is rubric-based and explainable. Your submission is scored against weighted criteria (the skill definition, progressive disclosure, the command, the permission set, trigger scoping, and the demonstration) with per-criterion feedback quoted from your file. The pass threshold is 65 percent and you can resubmit. These are the Claude Code extensibility skills the Claude Certified Architect exam tests.

Frequently asked questions

How is this different from the Claude Code config task?

That task sets the project baseline (CLAUDE.md, rules, a CI gate). This one is about extending Claude Code: authoring a Skill with progressive disclosure, a custom command, and a permission set, which is a different slice of the configuration domain.

What is progressive disclosure in a skill?

Keeping the skill body short and referencing bundled files or scripts for the heavy detail, so the model only loads the full detail when it actually runs the skill. It keeps skills cheap and composable.

Why a permission set?

A skill that can run any tool is a liability. Declaring allowed and denied tools and choosing the least-powerful permission mode that still works is how you let the agent automate a workflow without giving it free rein.

What counts as a complete submission?

A single file containing a SKILL.md with frontmatter and progressive disclosure, a .claude/commands command that takes arguments, a settings.json permission set with allowed/denied tools and a mode, stated trigger conditions, and a sample invocation.