Claude GuideDevantia × Executive Partners Group
DevantiaExecutive Partners Group
0/23 Jean-Christophe Leroy
Jean-Christophe Leroy
Guide author
✓ Verified on September 3, 2026⏱ 5 min read
Module 06

Claude Code — the AI terminal

In plain terms — who this module is for Claude Code is the tool for your technical teams (IT, data, developers). As an executive, you don't have to use it yourself: just remember that it lets Claude build and deploy custom tools (internal applications, automations, connections to your systems). What follows is deliberately more technical — skim it for background, or pass it on to your IT department.
npm install -g @anthropic-ai/claude-code cd my-project && claude # Useful commands /voice → voice mode | /compact → compact | /powerup → tutorials | /stats → tokens

🖥️ Desktop mode

Graphical interface, visual diff, app preview. No terminal.

☁️ Remote sessions

Cloud execution. Close your Mac, the work continues.

What an executive needs to know

Claude Code evolves fast. Even from a distance, here are the capabilities that matter to leadership:

🔁 Routines

Automations configured once (instructions + repository + connectors) then run on Anthropic's cloud, on a schedule or on a GitHub event. Monitoring, reports, and fixes run with no one at the screen.

🐝 Dynamic workflows (Opus 4.8 & 5)

Claude writes its own orchestration script that steers dozens of sub-agents in parallel — audits and migrations across an entire software estate.

🔎 /ultrareview

A fleet of "bug-hunter" agents runs in the cloud and feeds verified issues back into the session. As a complement: /code-review for a targeted pass.

🛡️ Claude Code Security

A security scan that reasons like a human researcher (tracing data flows between files) with adversarial verification to weed out false positives. Enterprise beta. Scans run on Mythos 5 — the first time this model is available outside Glasswing, in a locked-down flow: code in, findings report (CWE, severity) and proposed fixes out.

🧠 Automatic memory

Claude keeps its own persistent notes from one session to the next (MEMORY.md index), on top of the CLAUDE.md written by the team.

🌐 Built-in browser

The desktop app ships with a sandboxed browser: Claude reads docs, tests an interface, fills out a form — under guardrails (no purchases, no account creation, no CAPTCHA).

📱 Remote Control

A session running on a workstation can be steered from claude.ai or your phone: monitoring, sending files, approving steps remotely.

⚡ Fast mode & effort

/fast: Opus up to 2.5× faster (higher cost per token). /effort sets the depth of reasoning, up to the xhigh level for the hardest problems.

🗓️ Fable 5.1, restricted mode, costs

/model fable gives access to Fable 5.1 (v2.1.255+; it is not the default anywhere). --restricted starts a locked-down session (no command line, files limited to the working folder). /claude-api cost-optimize profiles your spend and suggests levers (cache, model, effort); /usage now breaks costs down by loop. New "Concise" output style and built-in spellcheck.

📌 Auto mode — the end of "allow everything" A security classifier replaces permission requests: safe actions go through, risky actions are blocked. It's the balance between approving everything by hand and granting full powers. Have your IT department enable this rather than the "skip permissions" mode.
🛠️ For IT — technical details (click to expand) · Anatomy of the .claude/ folder

Anatomy of the .claude/ folder

Section for your technical teams. Think of the .claude/ folder as a department's procedures binder: it brings together in one place the team's instructions, the rules to follow, and the automated tasks, so everyone works the same way. Every Claude Code project contains one. Here's its structure:

your-project/ ├── CLAUDE.md → Team instructions (committed to git) ├── CLAUDE.local.md → Your personal overrides (gitignored) └── .claude/ ├── settings.json → Permissions + config (committed) ├── settings.local.json→ Personal permissions (gitignored) ├── commands/ → Custom slash commands │ ├── review.md → /project:review │ ├── fix-issue.md → /project:fix-issue │ └── deploy.md → /project:deploy ├── rules/ → Modular instruction files │ ├── code-style.md │ ├── testing.md │ └── api-conventions.md ├── skills/ → Auto-invoked workflows │ ├── security-review/ │ └── deploy/ │ └── SKILL.md └── agents/ → Subagent personas (isolated) ├── code-reviewer.md └── security-auditor.md
📌 CLAUDE.md = permanent memory This is the most important file. Claude reads it automatically at the start of every session. Put your code conventions, your business rules, your non-negotiables in it. Commit it to git so the whole team benefits.
🛠️ For IT — technical details (click to expand) · The essential slash commands of Claude Code

The essential slash commands of Claude Code

Reference for your technical teams. Most users type a prompt and cross their fingers; advanced users steer Claude with these documented commands.

Session control

/costSee token consumption live
/compactCompress the chat, keep the essentials
/rewindGo back to an earlier point
/clearStart fresh, clutter-free
/resumeReopen an old session by its name
/branchCreate an alternate conversation branch
/exportSave the full chat as plain text
/renameName your session properly

Model & usage

/usageCheck plan limits instantly
/modelSwitch between Fable 5, Opus, Sonnet, and Haiku
/extra-usageKeep going when limits are reached
/fastSwitch Opus to fast mode (2.5×)
/effortSet the depth of reasoning (up to xhigh)

Project configuration

/initKick off a project with a CLAUDE.md guide
/add-dirAdd a working directory
/memoryEdit memory, turn auto-memory on/off
/scheduleCreate a Routine (cloud automation)
@fileReference a file (@report.csv) or a folder (@folder/)

Security & review

/security-reviewScan for vulnerabilities before shipping
/diffVisually see every uncommitted change
/planThink before building, make a plan first
/code-reviewCode review focused on bugs
/ultrareviewFleet of bug-hunting agents in the cloud

Agents & permissions

/agentsConfigure the sub-agent architecture
/permissionsControl exactly what Claude can touch
/skillsSee all available skills at once

Plugins & connectivity

/pluginManage Claude Code plugins
/reload-pluginsApply changes without restarting
/mcpManage all MCP server connections
/configOpen the full configuration
/doctorDiagnose installation problems
Esc EscReturn to the previous checkpoint
🎯 The 3 commands to pass on to your IT department /plan (Claude shows every step before acting — zero surprises), /security-review (the safety net before any deployment), and /cost (control of the token budget).
🛠️ Your turn — 5 minutes

Pass on to your CIO the three commands to remember (/plan, /security-review, /cost) and ask for a one-line reply.

Expected result : Your CIO knows what you expect from Claude Code: building under control.