⚠ Update (June 12, 2026): Fable 5 and Mythos 5 were suspended globally by US government order four days after their launch. All other models remain available. Fable 5 entries in the cheat sheet are preserved for reference, marked as suspended.
I republished a Claude Code cheatsheet at the end of April. Seven weeks later it has aged a lot: between v2.1.101 and v2.1.170, Anthropic shipped roughly sixty releases. Two new models landed (Opus 4.8 on May 28, then Fable 5 on June 9), the Agent View arrived, dynamic workflows now orchestrate hundreds of subagents, and a handful of new commands (/goal, /cd, /workflows, /ultrareview, /reload-skills…) showed up.
So I republished a June 2026 edition — same visual architecture as the March and April versions, rewritten to match every addition, using the blog palette (dark background, gold accent).
The cheat sheet
The previous editions stay available: April EN · FR, March EN · FR.
Items with a gold NEW badge shipped since the April 2026 edition.
What's covered
16 sections (two new cards: Agent View, and Dynamic Workflows & Security):
| Section | Coverage |
|---|---|
| Keyboard Shortcuts | ~30 shortcuts, transcript navigation, Vim mode |
| Slash Commands | 65+ commands in 6 categories |
| CLI Flags & Subcommands | 70+ flags in 9 categories, claude agents |
| The Big 5 (Extension System) | CLAUDE.md, Rules, Commands, Skills, Subagents, MCP, Plugins |
| Permission Modes | 6 modes + June hardening (startup files, build configs) |
| Hooks | 27 events, 5 handlers, MessageDisplay, args[] |
| Input Superpowers | @mentions, !shell, pipes, worktrees, --bg --exec |
| Configuration | 5-level hierarchy, env vars, settings.json |
| File Structure Map | Project + global tree |
| Rewind & Checkpoints | Rewind options + "Summarize up to here" |
| Auto Mode | Classifier, GA on Pro, hard deny, Bedrock/Vertex/Foundry |
| Models & Effort Levels | Fable 5, Opus 4.8/4.7, Sonnet 4.6, Haiku 4.5 + ultracode |
| Platforms | CLI (Windows without Git Bash), Desktop, Web (Routines), IDE |
| Agent View | claude agents, pinned sessions, background exec |
| Dynamic Workflows & Security | /workflows, security-guidance plugin, /ultrareview |
| Pro Workflow | Plan → Execute, /goal, fallback chains, safe mode |
| Customize Claude Code | Full frontmatter + plugin scaffolding |
| Quick Reference | Combos refreshed with June additions |
Additions and changes since April 2026
Claude Opus 4.8 (May 28)
The new default model on Max, Team Premium, Enterprise pay-as-you-go, and the Anthropic API (requires v2.1.154+). Default high effort, xhigh for the hardest tasks, native 1M context. Fast mode now runs on Opus 4.8 at $10/$50 per MTok (2x the standard rate for ~2.5x the speed); Opus 4.6 fast mode is deprecated.
Claude Fable 5 (June 9)
The Mythos-class model, the most capable available in Claude Code: 1M context, up to 128K output tokens, $10/$50 per MTok. Built for tasks larger than a single sitting — it sustains long autonomous sessions, investigates before acting, and verifies its own work. It is not the default model: /model fable (or the best alias). Thinking is always on there and can't be disabled.
One important quirk: Fable 5 runs with safety classifiers (cybersecurity, biology). A flagged request automatically reroutes to Opus (4.8 on the API, 4.7 on AWS) with a notice in the transcript. To check whether a customization (CLAUDE.md, a security repo…) trips the fallback on the very first request, launch claude --safe-mode.
Agent View (claude agents, May 11)
The big UX feature: a single screen for all your sessions — what's running, what's blocked on you, what's done. claude agents or ←← from a session. "Pinned" background sessions stay alive when idle, /resume lists them with their elapsed duration, and claude --bg --exec "pytest -x" launches an attachable shell job.
Dynamic workflows & ultracode (May 28)
A workflow is an orchestration script Claude writes for your task and runs across dozens to hundreds of subagents in the background — ideal for a codebase-wide audit or a large migration. Manage them with /workflows, and trigger them via the ultracode effort level (xhigh + auto orchestration) or the ultracode keyword (the former workflow, renamed early June).
The security-guidance plugin (May 28)
An official plugin that reviews Claude's changes for vulnerabilities and fixes them in the same session: a fast pattern check on each edit, a model review at the end of each turn, a deeper agentic review on commit/push. /plugin install security-guidance@claude-plugins-official then /reload-plugins. Project rules in .claude/claude-security-guidance.md.
New slash commands
/goal (set a completion condition, Claude iterates until it holds), /cd (change cwd without breaking the prompt cache), /scroll-speed, /workflows, /reload-skills, /ultrareview (fleet of cloud bug-hunting agents), /web-setup. The /code-review (correctness bugs, --comment/--fix) and /simplify (cleanup-only review that applies fixes) pair was re-split.
Auto Mode GA on Pro
The permission classifier now runs on Pro (with Sonnet 4.6 or Opus) and no longer requires opt-in consent. New: autoMode.hard_deny for rules that block unconditionally, and availability on Bedrock/Vertex/Foundry via CLAUDE_CODE_ENABLE_AUTO_MODE=1.
27th hook event + refinements
New MessageDisplay event (transform/hide displayed text). Plus: Stop/SubagentStop can return additionalContext, command hooks accept an exec form args: string[] (no more quote escaping), PostToolUse gains continueOnBlock, SessionStart can return reloadSkills: true, and $CLAUDE_EFFORT / effort.level are visible to hooks.
Fallback chains, safe mode & version gating
--fallback-model now accepts a chain (up to 3 models, comma-separated) and applies in interactive sessions too; persist it via fallbackModel (array) in settings. --safe-mode (and CLAUDE_CODE_SAFE_MODE) disables all customizations for debugging. Admins can enforce requiredMinimumVersion / requiredMaximumVersion, availableModels + enforceAvailableModels, and modelOverrides.
Platforms
Windows without Git Bash (PowerShell becomes the default shell when Bash is absent). Routines on Claude Code on the web (templated cloud agents fired by schedule, GitHub event, or API), mobile push notifications, custom themes, session recap. On the IDE side, Windsurf was renamed Devin Desktop.
Defaults that flipped
- Opus 4.8 default on Max / Team Premium / Enterprise pay-as-you-go / API
- Sonnet 4.6 on Pro / Team Standard / Enterprise subscription seats
- Opus 4.7 on Claude Platform on AWS (xhigh by default)
- Fable 5 never default (
/model fable) - Lean system prompt by default for all models except Haiku / Sonnet / Opus 4.7 and earlier
- Auto mode with no opt-in
Official sources: Claude Code Changelog, What's New, Model configuration, Dynamic workflows, Security guidance, Introducing Fable 5 & Mythos 5.
If you spot an error or a missing feature, please reach out. I'll keep this cheatsheet updated as Claude Code evolves.
