# Common Participant Questions

> **Summary — what this page covers**
> Quick answers to the questions that come up every time. Carry over the existing Day 1 FAQ and
> add the new steering-era ones. Keep answers to a few sentences.

## Should I use Claude Code or GitHub Copilot?

They do different jobs. Claude Code is an **agent** — it plans, reads whole-codebase context, runs
tools and commands, and loops until the task is done. Copilot is primarily **autocomplete and chat**
that helps *while you type*. They're not mutually exclusive: many developers keep Copilot for
in-the-editor completions and reach for Claude Code to execute whole tasks.

## Does Claude Code work with VS Code, Rider, and Visual Studio?

Claude Code runs in the **terminal** and integrates with **VS Code** and **JetBrains** (including
**Rider**, the .NET-native IDE) — it complements your editor rather than replacing it. The agent and
your `.claude/` config are identical across surfaces; only the diff/permission UI differs. There's no
dedicated **Visual Studio** extension today, so with VS run Claude Code in a terminal alongside it
(or use VS Code/Rider for the inline-diff experience). On Windows, run the terminal in **WSL2**.

## What happens when my context window fills up?

`/compact` summarizes the conversation and continues — but some detail is lost, so break big tasks
into smaller chunks and `/clear` between unrelated ones. Your **root CLAUDE.md and unscoped rules are
re-injected after compaction**, so your standing project briefing survives; ad-hoc conversation
detail may not.

## Is the code Claude generates safe to commit?

Review it like a **junior developer's PR** — you're still accountable for what you merge. Your
`PostToolUse` build/test hook ensures it at least compiles and passes tests, and a code-review skill
catches the common .NET failure modes (see [Section 4](09-section-4-sdlc.md)). Don't blind-merge.

## What does running Claude Code all day cost?

It needs a **paid plan** — it's not on Free. **Pro (~$20/mo)** is the entry tier and comfortably
covers a workshop day; **Max** is for heavy all-day use. Usage is a **shared, rolling 5-hour pool**
across Claude.ai and Claude Code. An **API key** is pay-as-you-go instead (the Day 2 path). See
[Before You Begin](01-before-you-begin.md#plan--access-requirements).

## Can Claude Code access my production database?

Not by default. Claude Code reads your **project files** and runs **approved commands** — it reaches
a database only through an **MCP server you explicitly configure**. The permission model plus a
`PreToolUse` guardrail hook keep it from touching anything you didn't authorize. Don't hand it
production credentials casually.

## (New) Aren't skills just markdown files?

No — a skill is a **folder** with a `SKILL.md` plus optional `references/` and `scripts/`. The
**folder is the unit**, not a single file. (This changed from the earlier single-file approach — see
[Section 2](05-section-2-steering.md).)

## (New) Rule vs CLAUDE.md — when do I use which?

**CLAUDE.md** holds the facts every session needs — keep it short, a map of the project.
**Rules** hold conventions, and **path-scoped** rules load only when you're working in the matching
files, so they don't cost context everywhere. Rule of thumb: facts → CLAUDE.md; coding conventions →
rules.

## (New) How do I actually enforce a "never"?

Not with a prompt — a CLAUDE.md "never do X" line is a suggestion the model *may* follow. Use a
**deterministic** guardrail: a `PreToolUse` **hook** (exit 2 to block), **permissions**, or
**managed settings** for non-overridable, org-wide enforcement.