Coding Fox codingfox Want to train your team?
Claude Code Field Guide updated May 30, 2026

Know what to
reach for next.

A current, sourced guide to the Claude Code features worth learning: what they do, when to use them, when not to, and one practice rep to make each one stick.

Claude Code 2.1.158 · Opus 4.8 117 items progress saves locally community counts are anonymous
NEW

Latest updates

If you already know the map, start here. This is the dated layer for meaningful changes since the guide was first published.

May 30, 2026

Opus 4.8 + Dynamic Workflows

/effort ultracode is not just "try harder." It is the mode where Claude can stop working turn-by-turn and write a Dynamic Workflow: an orchestration script that runs agent phases, passes results forward, and keeps the main session cleaner.

Mental model: subagents delegate work. Dynamic Workflows delegate the orchestration itself.
Claude Code 2.1.158

Auto mode widened beyond direct Anthropic API

Auto mode is now available on Bedrock, Vertex, and Foundry for Opus 4.7 and Opus 4.8 when enabled with CLAUDE_CODE_ENABLE_AUTO_MODE=1. That matters for teams whose real work runs through enterprise providers.

Use it carefully: this widens where unattended permission review can work; it does not remove the need for deny rules, sandboxing, or review.
Claude Code 2.1.157

Project-local plugin work got less ceremonial

Plugins in .claude/skills auto-load, claude plugin init <name> scaffolds local plugins, and plugin autocomplete improved. This makes repo-local team tooling feel less like publishing infrastructure.

Why it matters: repeated team workflows can live beside the repo before they deserve a marketplace or shared package.
May 28-29, 2026

Less tab switching, less accidental fan-out

Agent View can run attachable shell commands with ! command, worktrees can be entered mid-session, /chrome lets you pick the connected browser, /simplify is now a cleanup pass, and workflow keyword triggers can be disabled.

Why it matters: these are operator upgrades: fewer windows, cleaner cleanup, and fewer surprise expensive workflow launches.
Opus 4.8

Fast Mode is now a speed trade, not a mystery tax

Current changelog wording puts Opus 4.8 Fast Mode at about 2.5x the speed for 2x the standard rate. It is useful for live pairing and debugging, not a quality upgrade.

Rule: pay for speed only when waiting is the bottleneck. For background work, use normal mode and spend the saved budget on verification.
USE

How to use this

You are not supposed to read this front to back. Open the thing you need, read only as much as helps, try the rep on real work, and check it off only when it actually helped.

Click the square when you have used the item.
?Open the explanation, caveat, source, and practice idea.
Mark an item as worth knowing; community counts stay anonymous.
0 / 117 mastered
A

Skills & Plugins

A Skill is a SKILL.md file — markdown instructions plus YAML frontmatter — that loads when you type /name or when Claude matches its description. The whole game is progressive disclosure: only descriptions cost context at startup, bodies load on use, reference files load only when read.

Using Codex? The matching concepts are Codex skills, hooks, and plugins. Same basic split: procedures go in skills, guarantees go in hooks, and shared bundles become plugins. Open the Codex version →

A1The mental model

ThingRuns inContext costBest for
SkillYour conversation (or forks)Description always; body on invokeRepeatable procedures, on-demand reference
SubagentIsolated windowOnly the summary returnsVerbose work you don't want in context
CLAUDE.md ruleEvery turnEvery request, foreverFacts always true (stack, style, env)

If a CLAUDE.md section became a multi-step procedure instead of a fact, it should be a skill. If you keep pasting the same instructions, that's the signal.

A2SKILL.md anatomy

~/.claude/skills/fix-issue/SKILL.md
---
name: fix-issue
description: Fix a GitHub issue. Use when user says "fix issue", "resolve #N"...
disable-model-invocation: true   # side effects -> you fire it, not Claude
allowed-tools: Read Edit Bash(gh *) Bash(git *)
---
# Fix Issue $ARGUMENTS
!`gh issue view $ARGUMENTS`   # runs BEFORE Claude sees it, output injected inline
1. Read above. 2. Grep code. 3. Fix. 4. Test. 5. git commit -m "fix (#$ARGUMENTS)"
Gotcha: user-invocable: false only hides a skill from the / menu — Claude can still call it. To block Claude from auto-firing a destructive skill, use disable-model-invocation: true.

A3Plugins package it all

A plugin bundles skills + commands + agents + hooks + MCP servers + bin/ executables, shipped through a marketplace (a git repo with a marketplace.json). You get namespacing (/my-plugin:review), one-command install, and versioning for teammates. There's no built-in skill-creator — you author skills with Claude: have it write the SKILL.md, then test on fresh instances and refine.

⚡ try it now

Type /plugin, select a plugin to see its context cost, then /skills to toggle a noisy one to name-only.

B

Agents & Orchestration

Three primitives people conflate: subagents (isolated, return a summary), background agents (full sessions that survive your terminal closing), and agent teams (peers that message each other). Plus the claude agents dashboard to drive them all.

Using Codex? Think App threads, built-in/custom subagents, and app worktrees. Codex puts more of the project in a desktop app; Claude Code is more terminal/TUI shaped. Compare the Codex agent model →

B1Subagents

A fresh Claude session with its own context, restricted tools, and chosen model. It can read 50 files and only the answer lands back in your context. Defined in .claude/agents/*.md. Use model: haiku for search, isolation: worktree to keep it from colliding, memory: project to persist learnings across sessions.

B2Agent View & teams

Run claude agents for a full-screen dashboard of every running, blocked, and finished session — peek with Space, background any session with . Agent teams let a lead spawn teammates that talk to each other via a shared task list, not just back to the lead. Fire-and-forget with claude --bg "..." and keep working.

The catch: teams burn a lot more tokens — cost scales with how many teammates you run, and 3–5 is the practical starting range before coordination starts eating the win. And if two teammates edit the same file, one quietly overwrites the other. Give each its own files.
⚡ try it now

Run claude agents, then claude --bg "summarize this repo in 5 bullets", and hit Space on the row to peek. The whole background loop in 30 seconds.

C

CLAUDE.md & Memory

CLAUDE.md loads every session, every request — so every line is a recurring token tax. Anthropic's own docs put it bluntly: "Bloated CLAUDE.md files cause Claude to ignore your actual instructions." Target under 200 lines.

Using Codex? The equivalent durable instruction file is AGENTS.md, with global/project/nested discovery and a documented byte budget. See AGENTS.md →

C1The one-question test

For every line: "Would removing this cause a mistake?" If not, cut it. Keep commands Claude can't guess, style that differs from defaults, repo etiquette, env quirks. Drop standard conventions, "write clean code," and API docs (link them instead).

C2Hierarchy, imports & memory

Files concatenate root-down: managed → ~/.claude/CLAUDE.md → project → CLAUDE.local.md. Nested package files load on demand. Use @imports for shared conventions and .claude/rules/ with paths: for file-scoped rules. Separately, Auto Memory (on by default) lets Claude write its own MEMORY.md per repo. Auto Dream, which consolidates those notes, is still experimental and rolling out — don't count on having it yet.

⚡ try it now

Run /memory to see exactly what's loaded this session — and which files Claude is quietly writing to its own memory.

D

Hooks · MCP · Config

The enforcement layer. CLAUDE.md is a request; a hook is a deterministic gate you still review like code. MCP plugs in external tools. Sandbox and permissions decide what any of it can touch.

Using Codex? Codex also has MCP, hooks, config, and sandbox/approval concepts. Use MCP for live tools, hooks for deterministic gates, and config for safety posture. Open the Codex integration map →

D1Hooks

Shell commands that fire at lifecycle events (~30 of them). The daily ones: PostToolUse to auto-format on edit, PreToolUse to block edits to .env, Stop to run a test-gate that feeds failures back, Notification to ping you when Claude's idle. The new if field scopes a hook to specific calls.

D2MCP & Tool Search

Add servers with claude mcp add (http/stdio), three scopes, OAuth via /mcp. In current Claude Code builds, Tool Search defers tool definitions, so a big MCP roster is less likely to drown your context; check live docs if this behavior changes. A good starting set: GitHub + Context7 + Playwright.

D3Sandbox & permissions

OS-level sandboxing (Seatbelt / bubblewrap), paired with explicit deny rules for sensitive paths, can stop injected commands from reaching places like ~/.ssh. Pair it with autoAllowBashIfSandboxed to cut approval fatigue. Permission modes cycle with Shift+Tab; /fewer-permission-prompts builds you a safe allowlist from your own transcripts.

⚡ try it now

Run /sandbox for your isolation status, then /fewer-permission-prompts to auto-generate an allowlist and cut the interruptions.

E

The 2026 Drops

In four months Claude Code grew a cloud execution surface, scheduled agents, self-driving loops, and phone access. The ones actually worth knowing:

Using Codex? The comparable layer is Codex App automations, thread automations, cloud tasks, and mobile access. Same warning: scheduled work should produce useful artifacts, not unread transcripts. Compare Codex automations →
Heads up: the original Sonnet 4 & Opus 4 retire June 15, 2026. Current lineup: Opus 4.8, Sonnet 4.6, Haiku 4.5.
How to think about Ultracode

Subagents delegate work. Dynamic Workflows delegate the orchestration itself. Reach for /effort ultracode when the task needs independent lanes, cross-checking, or repeatable phases. Drop back to /effort high when the work is routine, sequential, or same-file.

⚡ try it now

Run /powerup — Anthropic's own in-terminal lessons. It exists because the tool ships weekly. It's the catch-up button.

F

Using Claude Code + Codex Together

The move is not "ask two AIs the same vague thing." The move is independent attempts, adversarial review, and one synthesis pass against the same acceptance criteria. Two strong agents are useful when they disagree for good reasons.

Using Codex? The Codex-side version is the same operating pattern with Codex threads, app worktrees, cloud tasks, and mobile access. Open the Codex version →
PatternHow to run itWhy it works
Alternative reviewsClaude writes, Codex reviews. Then flip it.Different failure modes. The reviewer is not emotionally attached to the diff.
Head-to-head buildGive both the same task in separate worktrees or branches.You get two concrete solutions instead of one confident guess.
Competition framingTell each one who it is competing against, then judge by evidence.It often raises effort, but it is a framing trick — tests and acceptance criteria still decide.
Synthesis passA fresh agent compares both outputs and chooses, merges, or rejects.Prevents "average the two answers" compromise mush.

Head-to-head builder prompt

Use this when Claude Code and Codex are each working in separate worktrees or branches.

You are competing against <competitor name>. Build the best solution to this exact task.

Task:
<paste the task>

Acceptance criteria:
<paste the checklist>

Rules:
- Same repo, same constraints, no shortcuts.
- Optimize for correctness, smallest safe diff, clear product fit, and verification evidence.
- Do not imitate the competitor. Beat them by being more precise.
- Done means: changed files listed, tests/checks run, risks named, and any manual verification described.

Adversarial review prompt

Use this when one tool produced a diff and the other gets a clean review pass.

Review this work as if <competitor name> already approved it and your job is to find what they missed.

Focus on:
- correctness and hidden edge cases
- security, data loss, privacy, and permission mistakes
- missing tests or weak verification
- unnecessary scope creep
- mobile/browser/product regressions

Give file/line evidence. Rank findings by severity. If there are no real issues, say that clearly and name the residual risk.

Synthesis prompt

Use this after two implementations or two reviews. Do not let it average both answers just to sound fair.

Compare Solution A and Solution B. Pick the winner or merge the best parts.

Judge by:
- correctness
- maintainability
- smallest safe diff
- test and browser evidence
- product fit
- risk of hidden follow-up work

Do not compromise by averaging. Make a decision, explain it, and give the next concrete action.
Practical test: competition is useful only when the verifier is boring and objective. Same task, same constraints, separate worktrees, fresh reviewer, evidence at the end.
G

Quick Reference

Type / in any session to see yours. The ones worth memorizing:

CommandDoesCommandDoes
/goalAuto-loop until a condition's met/agentsManage subagents
/loopRecurring scheduled task/mcpManage MCP servers
/powerupInteractive lessons/memoryAudit loaded context
/usageWhat's driving your limits/sandboxIsolation status
/doctorDiagnose config/budget/themeCustom colors
/rewindRoll back / summarize/fewer-permission-promptsAuto-allowlist

CLI: claude agents · claude --bg "..." · claude --worktree x · claude --agent X · claude mcp add · claude plugin validate.

rate this guide: