Module 03 — Advanced configuration

Name: Harness Engineering — Claude Code Mastery
Price: 89 EUR
Availability: InStock

In this module

You'll know how to configure Claude Code in depth: manage the 5 settings scopes, lock down permissions with allow/ask/deny, enable sandboxing, leverage automatic memory, and tune the model's reasoning effort.

Contents

3.1 — The five configuration scopes Priority hierarchy (highest to lowest) Merge rule Key takeaways

3.2 — settings.json in practice Create the three files Best practices Key takeaways 3.3 — The permissions system: allow, ask, deny The three lists Pattern syntax Recommended configuration Key takeaways 3.4 — Permission modes The six modes Activate a mode Multi-mode strategy Key takeaways 3.5 — Sandboxing Filesystem — restrict read and write Network — control accessible domains autoAllowBashIfSandboxed — the productivity/security trade-off Key takeaways 3.6 — Automatic memory Architecture File locations Enable and manage Memory types Key takeaways 3.7 — Choosing and configuring your model The three models Configuration Effort levels — tune the thinking Extended thinking and /ultrathink Key takeaways 3.8 — Environment variables and personalization Environment variables (env) Language and style Other useful settings Full User file example Key takeaways 3.9 — Team configuration: Project and Managed The Project scope — team convention The Managed scope — enterprise governance CLAUDE.md + settings.json = full combo Key takeaways 3.10 — Debugging your configuration with /status What /status shows Debugging methodology Common issues Key takeaways

3.1 — The five configuration scopes Priority hierarchy (highest to lowest) Merge rule Key takeaways

3.1 — The five configuration scopes

Claude Code doesn't read a single config file. It merges up to five layers, each with a different priority level. Understanding this hierarchy lets you predict exactly which setting will be active in any situation.

Priority hierarchy (highest to lowest)

Priority	Scope	File / Source	Shared?
1	Managed	`managed-settings.json` or `managed-settings.d/`	Enterprise (MDM)
2	CLI	Flags `--model`, `--permission-mode`, etc.	Session only
3	Local	`.claude/settings.local.json` (project root)	No (gitignored)
4	Project	`.claude/settings.json` (project root)	Yes (git-tracked)
5	User	`~/.claude/settings.json`	No (global personal)

Merge rule

Scalar values (strings, booleans) are overwritten by the highest-priority scope. Arrays merge across scopes. Concretely, if the Project scope defines permissions.allow: ["Edit(*.kt)"] and the User scope defines permissions.allow: ["Bash(git status)"], the final result contains both entries.

Claude Code

# Check the final merge result at any time
/status

Key takeaways

• 5 scopes: Managed > CLI > Local > Project > User

• Scalars overwrite, arrays merge

• .claude/settings.json = team, .claude/settings.local.json = personal

• /status to inspect the effective configuration

In this module

3.1 — The five configuration scopes Available
3.2 — settings.json in practice
3.3 — The permissions system: allow, ask, deny
3.4 — Permission modes
3.5 — Sandboxing
3.6 — Automatic memory
3.7 — Choosing and configuring your model
3.8 — Environment variables and personalization
3.9 — Team configuration: Project and Managed
3.10 — Debugging your configuration with /status

Locked content

Unlock the full course to access this module.

Unlock

Already purchased? Recover access