docs: sync to Claude Code v2.1.212 (#154)

* docs: sync to Claude Code v2.1.212 Bumps tutorial coverage from the v2.1.206 baseline (2026-07-11 sync) to v2.1.212, plus a repo-internal accuracy audit that surfaced defects independent of the version delta. Fixed: - Dead `#` memory-shortcut removed from 02-memory/README.md (two places documented a discontinued quick-memory-add pattern as working, directly contradicting the file's own commands table) - Auto mode opt-in -> opt-out on Bedrock/Vertex/Foundry (v2.1.207) - `auto` permission mode mislabeled "Research Preview" across 09-advanced-features/README.md, CATALOG.md, README.md, and the vi/ja/zh/uk translations (auto mode is GA, gated only by plan/model/provider) - `/fork` / `/branch` history corrected: they were a single aliased command only from v2.1.77 to v2.1.161, not "renamed" as previously stated - `effort` frontmatter enum in 04-subagents/README.md completed with `xhigh` - Bundled-skills count reconciled in CATALOG.md (9 -> 10) - INDEX.md Skills section count and Feature Coverage Matrix Total column fully recomputed and reconciled (Skills: 22, Plugins: 39) - Stale 2.1.160 footer cluster (5 files) bumped to 2.1.212 - QUICK_REFERENCE.md fully resynced (was 52 versions behind) - 02-memory/README.md's Memory Hierarchy section rewritten: replaced a fabricated 8-tier strict-precedence model with the verified structure (CLAUDE.md files are concatenated into context, not overridden; 4 real location tiers; rules and auto memory are separate mechanisms) - 02-memory/README.md's Memory Architecture diagram no longer conflates claude.ai's 24h synthesis cycle with Claude Code's continuous auto memory - Import recursion depth corrected (was documented as 5, actually 4 hops) Added: - Subagent output scanning (v2.1.210) documented in 04-subagents/README.md - Session-wide spawn caps (v2.1.212): WebSearch and subagent limits - MCP long-running-tool auto-backgrounding (v2.1.212) in 05-mcp/README.md - `claude auto-mode reset` and `/resume` picker (v2.1.212) - Screen reader mode (v2.1.208): --ax-screen-reader / CLAUDE_AX_SCREEN_READER - Task tool `mode` parameter deprecation note (v2.1.212) Known gaps (deferred, not fixed this pass): - 03-skills/blog-draft/ vs 03-skills/.claude/skills/blog-draft/: confirmed the latter is gitignored local testing scratch, not a repo duplicate * fix(INDEX): correct Skills row arithmetic and stale changelog notes INDEX.md's Skills header/matrix shipped as 22 (5+9+8) despite the PR's own CHANGELOG claiming it was fixed to 21 — three independent counts (file-type breakdown, filesystem listing, changelog intent) agree on 21; the Docs column was off by one. Also removes two stale CHANGELOG "Known gaps" entries this same PR already resolved (memory hierarchy rewrite, blog-draft gitignore confirmation) and adds the missing Fixed-section bullet documenting the memory hierarchy rewrite.
2026-07-20 23:30:51 +02:00 · 2026-07-18 22:21:21 +07:00
parent a645ffe6b0
commit 8f045173d1
24 changed files with 336 additions and 211 deletions
@@ -24,7 +24,8 @@ Built-in commands are shortcuts for common actions. There are **60+ built-in com
 |---------|---------|
 | `/add-dir <path>` | Add working directory |
 | `/agents` | Manage agent configurations |
-| `/branch [name]` | Branch conversation into a new session (alias: `/fork`). Note: `/fork` renamed to `/branch` in v2.1.77 |
+| `/branch [name]` | Switch into a copy of the conversation at this point, preserving the original (return to it with `/resume`) |
+| `/fork <directive>` | Spawn a background subagent that inherits the full conversation and works on the directive while you keep going; its own row in `claude agents` |
 | `/btw <question>` | Ask an ephemeral side question while Claude is working on the main task; doesn't pollute the main conversation context |
 | `/cd <path>` | Move the session to a new working directory without breaking the prompt cache (added v2.1.169) |
 | `/chrome` | Configure Chrome browser integration |
@@ -121,13 +122,13 @@ These skills ship with Claude Code and are invoked like slash commands:
 | Command | Status |
 |---------|--------|
 | `/output-style` | Deprecated since v2.1.73 |
-| `/fork` | Renamed to `/branch` (alias still works, v2.1.77) |
 | `/pr-comments` | Removed in v2.1.91 — ask Claude directly to view PR comments |
 | `/vim` | Removed in v2.1.92 — use /config → Editor mode |

 ### Recent Changes

- `/fork` renamed to `/branch` with `/fork` kept as alias (v2.1.77)
+- `/fork` was an alias for `/branch` from v2.1.77 until v2.1.161, when they became distinct commands: `/fork` now spawns a background subagent that inherits the conversation, while `/branch` switches you into a copy in place
+- `/resume` (no arguments) opens a picker of past sessions — including ones removed from the visible list — and resumes the chosen one as a background session (v2.1.212)
 - `/output-style` deprecated (v2.1.73)
 - `/review <pr>` now uses the same review engine as `/code-review medium` (v2.1.186)
 - `/effort` command added; `max` level available on Opus 4.6+ (originally Opus 4.6-only)
@@ -632,8 +633,8 @@ If both exist with the same name, the **skill takes precedence**. Remove one or

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/slash-commands
 - https://code.claude.com/docs/en/skills
@@ -183,7 +183,7 @@ See @docs/architecture.md for system design
 **Import features:**

 - Both relative and absolute paths are supported (e.g., `@docs/api.md` or `@~/.claude/my-project-instructions.md`)
- Recursive imports are supported with a maximum depth of 5
+- Recursive imports are supported with a maximum depth of 4 hops
 - First-time imports from external locations trigger an approval dialog for security
 - Import directives are not evaluated inside markdown code spans or code blocks (so documenting them in examples is safe)
 - Helps avoid duplication by referencing existing documentation
@@ -191,85 +191,68 @@ See @docs/architecture.md for system design

 ## Memory Architecture

-Memory in Claude Code follows a hierarchical system where different scopes serve different purposes:
+Memory in Claude Code follows a hierarchical system where different scopes serve different purposes. Unlike Claude Web/Desktop's 24-hour synthesis cycle (see [Memory in Claude Web/Desktop](#memory-in-claude-webdesktop) below), Claude Code has two memory systems that both load at the start of every session and update continuously, not on a timer:

 ```mermaid
 graph TB
-    A["Claude Session"]
-    B["User Input"]
-    C["Memory System"]
-    D["Memory Storage"]
+    A["Session Start"]
+    B["CLAUDE.md Files<br/>(you write)"]
+    C["Auto Memory<br/>(Claude writes)"]
+    D["Claude Session"]
+    E["Your Correction /<br/>Preference"]

-    B -->|User provides info| C
-    C -->|Synthesizes every 24h| D
-    D -->|Loads automatically| A
-    A -->|Uses context| C
+    B -->|loaded in full| A
+    C -->|MEMORY.md loaded| A
+    A --> D
+    D -->|"Remember that..."| E
+    E -->|writes during session| C
+    D -->|"add this to CLAUDE.md"| B
 ```

 ## Memory Hierarchy in Claude Code

-Claude Code uses a multi-tier hierarchical memory system. Memory files are automatically loaded when Claude Code launches, with higher-level files taking precedence.
+Claude Code has two complementary memory systems, both loaded at the start of every conversation: **CLAUDE.md files** (instructions you write) and **auto memory** (notes Claude writes itself). CLAUDE.md files are **concatenated into context rather than overriding each other** — this is not a strict precedence chain where a higher tier replaces a lower one. `.claude/rules/*.md` files are a separate, related mechanism for topic- or path-scoped instructions.

-**Complete Memory Hierarchy (in order of precedence):**
+**CLAUDE.md file locations, in load order (broadest scope to most specific):**

-1. **Managed Policy** - Organization-wide instructions
-   - macOS: `/Library/Application Support/ClaudeCode/CLAUDE.md`
-   - Linux/WSL: `/etc/claude-code/CLAUDE.md`
-   - Windows: `C:\Program Files\ClaudeCode\CLAUDE.md`
+| Scope | Location | Purpose |
+|-------|----------|---------|
+| Managed policy | macOS: `/Library/Application Support/ClaudeCode/CLAUDE.md`<br>Linux/WSL: `/etc/claude-code/CLAUDE.md`<br>Windows: `C:\Program Files\ClaudeCode\CLAUDE.md` | Organization-wide instructions managed by IT/DevOps. Cannot be excluded by individual settings. |
+| User instructions | `~/.claude/CLAUDE.md` | Personal preferences for all projects |
+| Project instructions | `./CLAUDE.md` or `./.claude/CLAUDE.md` | Team-shared instructions, version controlled |
+| Local instructions | `./CLAUDE.local.md` | Personal project-specific preferences; add to `.gitignore` |

-2. **Managed Drop-ins** - Alphabetically merged policy files (v2.1.83+)
-   - `managed-settings.d/` directory alongside the managed policy CLAUDE.md
-   - Files are merged in alphabetical order for modular policy management
+Within the directory tree, Claude Code walks up from your working directory: `foo/CLAUDE.md` loads before `foo/bar/CLAUDE.md` if you launch from `foo/bar/`, so instructions closer to where you launched are read *last* — not "highest priority" in an override sense, just most recent in context. Within each directory, `CLAUDE.local.md` is appended after `CLAUDE.md`. CLAUDE.md and CLAUDE.local.md files in subdirectories *under* your working directory load on demand, when Claude reads files in those subdirectories, rather than at launch.

-3. **Project Memory** - Team-shared context (version controlled)
-   - `./.claude/CLAUDE.md` or `./CLAUDE.md` (in repository root)
+Organizations can also put managed CLAUDE.md content directly inside `managed-settings.json` via the `claudeMd` key, instead of deploying a separate file. This is honored only in managed/policy settings — setting `claudeMd` in user or project settings has no effect.

-4. **Project Rules** - Modular, topic-specific project instructions
-   - `./.claude/rules/*.md`
+**`.claude/rules/*.md`** — modular, topic-specific instructions, optionally scoped to file paths via `paths` frontmatter. Rules without a `paths` field load unconditionally with the same priority as `.claude/CLAUDE.md`; path-scoped rules load on demand when Claude reads a matching file. User-level rules (`~/.claude/rules/`) load before project rules.

-5. **User Memory** - Personal preferences (all projects)
-   - `~/.claude/CLAUDE.md`
-
-6. **User-Level Rules** - Personal rules (all projects)
-   - `~/.claude/rules/*.md`
-
-7. **Local Project Memory** - Personal project-specific preferences
-   - `./CLAUDE.local.md`
+**Auto memory** (`~/.claude/projects/<project>/memory/`) is a separate system: Claude's own notes, not CLAUDE.md content, and not part of the concatenation order above. See [Auto Memory](#auto-memory) below.

 > **Note**: `CLAUDE.local.md` is fully supported and documented in the [official documentation](https://code.claude.com/docs/en/memory). It provides personal project-specific preferences that are not committed to version control. Add `CLAUDE.local.md` to your `.gitignore`.

-8. **Auto Memory** - Claude's automatic notes and learnings
-   - `~/.claude/projects/<project>/memory/`
-
 **Memory Discovery Behavior:**

-Claude searches for memory files in this order, with earlier locations taking precedence:
-
 ```mermaid
 graph TD
-    A["Managed Policy<br/>/Library/.../ClaudeCode/CLAUDE.md"] -->|highest priority| A2["Managed Drop-ins<br/>managed-settings.d/"]
-    A2 --> B["Project Memory<br/>./CLAUDE.md"]
-    B --> C["Project Rules<br/>./.claude/rules/*.md"]
-    C --> D["User Memory<br/>~/.claude/CLAUDE.md"]
-    D --> E["User Rules<br/>~/.claude/rules/*.md"]
-    E --> F["Local Project Memory<br/>./CLAUDE.local.md"]
-    F --> G["Auto Memory<br/>~/.claude/projects/.../memory/"]
+    A["Managed Policy<br/>/Library/.../ClaudeCode/CLAUDE.md"] -->|loads first| B["User Instructions<br/>~/.claude/CLAUDE.md"]
+    B --> C["Project Instructions<br/>./CLAUDE.md or ./.claude/CLAUDE.md"]
+    C --> D["Local Instructions<br/>./CLAUDE.local.md"]

-    B -->|imports| H["@docs/architecture.md"]
+    C -->|imports| H["@docs/architecture.md"]
    H -->|imports| I["@docs/api-standards.md"]

    style A fill:#fce4ec,stroke:#333,color:#333
-    style A2 fill:#fce4ec,stroke:#333,color:#333
-    style B fill:#e1f5fe,stroke:#333,color:#333
+    style B fill:#f3e5f5,stroke:#333,color:#333
    style C fill:#e1f5fe,stroke:#333,color:#333
-    style D fill:#f3e5f5,stroke:#333,color:#333
-    style E fill:#f3e5f5,stroke:#333,color:#333
-    style F fill:#e8f5e9,stroke:#333,color:#333
-    style G fill:#fff3e0,stroke:#333,color:#333
+    style D fill:#e8f5e9,stroke:#333,color:#333
    style H fill:#e1f5fe,stroke:#333,color:#333
    style I fill:#e1f5fe,stroke:#333,color:#333
 ```

+All files shown are concatenated into one context, not selected by override — later boxes appear later in context, not "instead of" earlier ones.
+
 ## Excluding CLAUDE.md Files with `claudeMdExcludes`

 In large monorepos, some CLAUDE.md files may be irrelevant to your current work. The `claudeMdExcludes` setting lets you skip specific CLAUDE.md files so they are not loaded into context:
@@ -292,16 +275,20 @@ Patterns are matched against paths relative to the project root. This is particu

 ## Settings File Hierarchy

-Claude Code settings (including `autoMemoryDirectory`, `claudeMdExcludes`, and other configuration) are resolved from a five-level hierarchy, with higher levels taking precedence:
+Claude Code settings (including `autoMemoryDirectory`, `claudeMdExcludes`, and other configuration) resolve by precedence — unlike CLAUDE.md files above, settings genuinely override rather than concatenate. When the same setting appears in multiple scopes, the higher level wins:

 | Level | Location | Scope |
 |-------|----------|-------|
-| 1 (Highest) | Managed policy (system-level) | Organization-wide enforcement |
-| 2 | `managed-settings.d/` (v2.1.83+) | Modular policy drop-ins, merged alphabetically |
+| 1 (Highest) | Managed — `managed-settings.json`, plist/registry, or server-managed | Organization-wide enforcement; cannot be overridden |
+| 2 | Command line arguments | Temporary session overrides |
 | 3 | `.claude/settings.local.json` | Local overrides (git-ignored) |
 | 4 | `.claude/settings.json` | Project-level (committed to git) |
 | 5 (Lowest) | `~/.claude/settings.json` | User preferences |

+Managed settings also support a drop-in directory, `managed-settings.d/`, alongside `managed-settings.json`: the base file merges first, then `*.json` files in the drop-in directory merge on top in alphabetical order (scalars override, arrays concatenate and de-duplicate, objects deep-merge). This lets separate teams deploy independent policy fragments without editing a shared file. Note this is a **settings.json** mechanism, not a CLAUDE.md one — it doesn't apply to the CLAUDE.md file locations described above.
+
+Permission rules (`allow`/`ask`/`deny`) behave differently from other settings: they merge across scopes instead of the higher level replacing the lower one.
+
 **Platform-specific configuration (v2.1.51+):**

 Settings can also be configured via:
@@ -416,18 +403,19 @@ Rules in `.claude/rules/` support two organizational features:

 ## Memory Locations Table

-| Location | Scope | Priority | Shared | Access | Best For |
-|----------|-------|----------|--------|--------|----------|
-| `/Library/Application Support/ClaudeCode/CLAUDE.md` (macOS) | Managed Policy | 1 (Highest) | Organization | System | Company-wide policies |
-| `/etc/claude-code/CLAUDE.md` (Linux/WSL) | Managed Policy | 1 (Highest) | Organization | System | Organization standards |
-| `C:\Program Files\ClaudeCode\CLAUDE.md` (Windows) | Managed Policy | 1 (Highest) | Organization | System | Corporate guidelines |
-| `managed-settings.d/*.md` (alongside policy) | Managed Drop-ins | 1.5 | Organization | System | Modular policy files (v2.1.83+) |
-| `./CLAUDE.md` or `./.claude/CLAUDE.md` | Project Memory | 2 | Team | Git | Team standards, shared architecture |
-| `./.claude/rules/*.md` | Project Rules | 3 | Team | Git | Path-specific, modular rules |
-| `~/.claude/CLAUDE.md` | User Memory | 4 | Individual | Filesystem | Personal preferences (all projects) |
-| `~/.claude/rules/*.md` | User Rules | 5 | Individual | Filesystem | Personal rules (all projects) |
-| `./CLAUDE.local.md` | Project Local | 6 | Individual | Git (ignored) | Personal project-specific preferences |
-| `~/.claude/projects/<project>/memory/` | Auto Memory | 7 (Lowest) | Individual | Filesystem | Claude's automatic notes and learnings |
+CLAUDE.md files and rules are concatenated into context, not selected by strict override — "Load order" below means *where in context it appears*, not *which one wins*. Auto memory is a separate mechanism with its own storage location.
+
+| Location | Type | Load order | Shared | Access | Best For |
+|----------|------|-------------|--------|--------|----------|
+| `/Library/Application Support/ClaudeCode/CLAUDE.md` (macOS) | Managed Policy | 1st (loads first) | Organization | System | Company-wide policies |
+| `/etc/claude-code/CLAUDE.md` (Linux/WSL) | Managed Policy | 1st (loads first) | Organization | System | Organization standards |
+| `C:\Program Files\ClaudeCode\CLAUDE.md` (Windows) | Managed Policy | 1st (loads first) | Organization | System | Corporate guidelines |
+| `~/.claude/rules/*.md` | User Rules | 2nd | Individual | Filesystem | Personal rules (all projects) |
+| `~/.claude/CLAUDE.md` | User Memory | 3rd | Individual | Filesystem | Personal preferences (all projects) |
+| `./.claude/rules/*.md` | Project Rules | 4th | Team | Git | Path-specific, modular rules |
+| `./CLAUDE.md` or `./.claude/CLAUDE.md` | Project Memory | 5th | Team | Git | Team standards, shared architecture |
+| `./CLAUDE.local.md` | Project Local | 6th (loads last) | Individual | Git (ignored) | Personal project-specific preferences |
+| `~/.claude/projects/<project>/memory/` | Auto Memory | N/A — separate mechanism | Individual | Filesystem | Claude's automatic notes and learnings |

 ## Memory Update Lifecycle

@@ -819,9 +807,7 @@ Claude has not save the rule because I did not have any `Claude.md` file anywher

 ### Example 4: Memory Update During Session

-You can add new rules to memory during an active Claude Code session. There are two ways to do this:
-
-#### Method 1: Direct Request
+You can add new rules to memory during an active Claude Code session by asking conversationally:

 ```markdown
 User: Remember that I prefer using React hooks instead of class components
@@ -846,30 +832,7 @@ Added to ./CLAUDE.md:
 - Use useMemo for expensive computations
 ```

-#### Method 2: Using "# new rule into memory" Pattern
-
-You can also use the explicit pattern "# new rule into memory" to trigger memory updates:
-
-```markdown
-User: # new rule into memory
-      Always validate user input with Zod schemas before processing.
-
-Claude: I'm saving this rule to memory. Which memory file should I use?
-        1. Project memory (./CLAUDE.md)
-        2. Personal memory (~/.claude/CLAUDE.md)
-
-User: Project memory
-
-Claude: ✅ Memory saved!
-
-Added to ./CLAUDE.md:
---
-
-### Input Validation
- Always validate user input with Zod schemas before processing
- Define schemas at the top of each API handler file
- Return 400 status with validation errors
-```
+Or use `/memory` to edit memory files directly for extensive updates or reorganization.

 #### Tips for Adding Memory

@@ -1070,20 +1033,6 @@ If you prefer manual setup:
   git commit -m "Add project memory configuration"
   ```

-#### Method 3: Quick Updates with `#`
-
-Once CLAUDE.md exists, add rules quickly during conversations:
-
-```markdown
-# Use semantic versioning for all releases
-
-# Always run tests before committing
-
-# Prefer composition over inheritance
-```
-
-Claude will prompt you to choose which memory file to update.
-
 ### Setup Personal Memory

 1. **Create ~/.claude directory:**
@@ -1170,22 +1119,22 @@ For the most up-to-date information, refer to the official Claude Code documenta
 **Import Syntax:**

 - Use `@path/to/file` to include external content (e.g., `@~/.claude/my-project-instructions.md`)
- Supports both relative and absolute paths
- Recursive imports supported with a maximum depth of 5
+- Supports both relative and absolute paths (relative paths resolve relative to the file containing the import, not the working directory)
+- Recursive imports supported with a maximum depth of 4 hops
 - First-time external imports trigger an approval dialog
 - Not evaluated inside markdown code spans or code blocks
 - Automatically includes referenced content in Claude's context

-**Memory Hierarchy Precedence:**
+**CLAUDE.md Load Order** (concatenated into context, not strict override — see [Memory Hierarchy in Claude Code](#memory-hierarchy-in-claude-code) above):

-1. Managed Policy (highest precedence)
-2. Managed Drop-ins (`managed-settings.d/`, v2.1.83+)
-3. Project Memory
+1. Managed Policy (loads first)
+2. User-Level Rules (`~/.claude/rules/`)
+3. User Memory
 4. Project Rules (`.claude/rules/`)
-5. User Memory
-6. User-Level Rules (`~/.claude/rules/`)
-7. Local Project Memory
-8. Auto Memory (lowest precedence)
+5. Project Memory
+6. Local Project Memory (loads last)
+
+Auto Memory is a separate mechanism (`~/.claude/projects/<project>/memory/`), not part of this concatenation order.

 ## Related Concepts Links

@@ -1199,8 +1148,8 @@ For the most up-to-date information, refer to the official Claude Code documenta
 - [Official Memory Docs](https://code.claude.com/docs/en/memory) - Anthropic documentation

 ---
-**Last Updated**: June 24, 2026
-**Claude Code Version**: 2.1.187
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/memory
 - https://code.claude.com/docs/en/settings
@@ -1210,4 +1159,4 @@ For the most up-to-date information, refer to the official Claude Code documenta
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.117
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.144
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.145
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -101,7 +101,7 @@ skills: skill1, skill2  # Optional - skills to preload into context
 mcpServers: server1  # Optional - MCP servers to make available
 memory: user  # Optional - persistent memory scope (user, project, local)
 background: false  # Optional - run as background task
-effort: high  # Optional - reasoning effort (low, medium, high, max)
+effort: high  # Optional - reasoning effort (low, medium, high, xhigh, max)
 isolation: worktree  # Optional - git worktree isolation
 initialPrompt: "Start by analyzing the codebase"  # Optional - auto-submitted first turn
 hooks:  # Optional - component-scoped hooks
@@ -126,14 +126,14 @@ to solving problems.
 | `tools` | No | Comma-separated list of specific tools. Omit to inherit all tools. Supports `Agent(agent_name)` syntax to restrict spawnable subagents |
 | `disallowedTools` | No | Comma-separated list of tools the subagent must not use |
 | `model` | No | Model to use: `sonnet`, `opus`, `haiku`, full model ID, or `inherit`. Defaults to configured subagent model |
-| `permissionMode` | No | `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan` |
+| `permissionMode` | No | `default`, `acceptEdits`, `dontAsk`, `bypassPermissions`, `plan`. As of v2.1.212, the Task tool's `mode` invocation parameter is deprecated and ignored — subagents inherit the parent session's permission mode by default unless overridden here |
 | `maxTurns` | No | Maximum number of agentic turns the subagent can take |
 | `skills` | No | Comma-separated list of skills to preload. Injects full skill content into the subagent's context at startup. **v2.1.133+:** subagents also discover project, user, and plugin skills via the Skill tool — same catalog as the main session, no longer limited to their own embedded set. |
 | `mcpServers` | No | MCP servers to make available to the subagent |
 | `hooks` | No | Component-scoped hooks (PreToolUse, PostToolUse, Stop) |
 | `memory` | No | Persistent memory directory scope: `user`, `project`, or `local` |
 | `background` | No | Subagents already run in the background by default (v2.1.198). Set to `true` to *force* background always and prevent inline execution |
-| `effort` | No | Reasoning effort level: `low`, `medium`, `high`, or `max` |
+| `effort` | No | Reasoning effort level: `low`, `medium`, `high`, `xhigh`, or `max`. Overrides the session effort level; available levels depend on the model |
 | `isolation` | No | Set to `worktree` to give the subagent its own git worktree |
 | `initialPrompt` | No | Auto-submitted first turn when the subagent runs as the main agent |

@@ -822,6 +822,20 @@ Plugin-provided subagents have restricted frontmatter capabilities for security.

 This prevents plugins from escalating privileges or executing arbitrary commands through subagent hooks.

+### Subagent Output Scanning (v2.1.210+)
+
+As of v2.1.210, Claude Code scans each subagent's final report for text that imitates the harness's own output format — fake `<system-reminder>`-style tags, fabricated `Human:`/`Assistant:` turns, or mentions of permission-bypass flags and settings-file paths. This defends against prompt injection carried in subagent output, such as a subagent that fetched a malicious web page containing fake control tokens designed to manipulate the parent session.
+
+When the scan flags something, Claude Code neutralizes it — inserting a backslash or an inline marker such as `[harness: subagent output matched instruction-shaped pattern(s): ...]` naming what triggered the scan — and the parent session is expected to treat the flagged text as a finding to relay, not an instruction to follow. The scan is on by default with no documented opt-out. It errs toward flagging: a legitimate subagent report that quotes a real flag name (e.g. `--dangerously-skip-permissions`) verbatim can trigger a marker even though nothing malicious occurred — a false positive is preferable to a missed injection.
+
+### Session-Wide Subagent Limit (v2.1.212+)
+
+Claude Code caps subagent spawns at **200 per session** by default, to stop runaway delegation loops. Override with `CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION`; the budget resets when you run `/clear`.
+
+```bash
+export CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION=200
+```
+
 ---

 ## Architecture
@@ -1246,8 +1260,8 @@ See the OpenTelemetry section in [Advanced Features → Telemetry](../09-advance

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/subagents
 - https://code.claude.com/docs/en/cli-reference
@@ -803,6 +803,15 @@ The maximum output limit is configurable via the `MAX_MCP_OUTPUT_TOKENS` environ
 export MAX_MCP_OUTPUT_TOKENS=50000
 ```

+## Auto-Backgrounding Long-Running Tool Calls (v2.1.212)
+
+MCP tool calls that run longer than 2 minutes now move to the background automatically, so the session stays usable instead of blocking on a slow tool. The threshold is configurable, and the behavior can be tuned or disabled, via `CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS`:
+
+```bash
+# Change the auto-background threshold to 5 minutes (300,000ms)
+export CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS=300000
+```
+
 ## Solving Context Bloat with Code Execution

 As MCP adoption scales, connecting to dozens of servers with hundreds or thousands of tools creates a significant challenge: **context bloat**. This is arguably the biggest problem with MCP at scale, and Anthropic's engineering team has proposed an elegant solution — using code execution instead of direct tool calls.
@@ -1187,8 +1196,8 @@ export GITHUB_TOKEN="your_token"

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/mcp
 - https://code.claude.com/docs/en/changelog
@@ -108,10 +108,10 @@ Result:

 ---

-**Last Updated**: June 2, 2026
-**Claude Code Version**: 2.1.160
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/plugins
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.131
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.138
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -120,10 +120,10 @@ export GITHUB_TOKEN="your_github_token"

 ---

-**Last Updated**: June 2, 2026
-**Claude Code Version**: 2.1.160
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/plugins
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.131
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.138
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -92,10 +92,10 @@ Result:

 ---

-**Last Updated**: June 2, 2026
-**Claude Code Version**: 2.1.160
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/plugins
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.131
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.138
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -339,9 +339,9 @@ User: Now let's continue with the approach that worked.
 6. **Summarize long sessions**: Use "Summarize from here" to keep conversations manageable

 ---
-**Last Updated**: June 2, 2026
-**Claude Code Version**: 2.1.160
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/checkpointing
 - https://code.claude.com/docs/en/changelog
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -48,7 +48,7 @@ Advanced features in Claude Code extend the core capabilities with planning, rea
 **Key advanced features include:**
 - **Planning Mode**: Create detailed implementation plans before coding
 - **Extended Thinking**: Deep reasoning for complex problems
- **Auto Mode**: Background safety classifier reviews each action before execution (Research Preview)
+- **Auto Mode**: Background safety classifier reviews each action before execution
 - **Background Tasks**: Run long operations without blocking the conversation
 - **Permission Modes**: Control what Claude can do (`manual` — formerly `default`, `acceptEdits`, `plan`, `auto`, `dontAsk`, `bypassPermissions`)
 - **Print Mode**: Run Claude Code non-interactively for automation and CI/CD (`claude -p`)
@@ -409,7 +409,7 @@ Toggle during a session with `Alt+T` / `Option+T`, set effort with `/effort`, or

 ## Auto Mode

-Auto Mode is a Research Preview permission mode (March 2026) that uses a background safety classifier to review each action before execution. It allows Claude to work autonomously while blocking dangerous operations.
+Auto Mode is a permission mode that uses a background safety classifier to review each action before execution. It allows Claude to work autonomously while blocking dangerous operations. It's available on all plans, but requires an eligible model (Claude Opus 4.6+, Sonnet 4.6+, or Fable 5 on the Anthropic API and Claude Platform on AWS; Sonnet 5, Opus 4.7, Opus 4.8, or Fable 5 on Bedrock, Vertex, Foundry, and signed-in Claude apps gateway sessions) and, on Team/Enterprise, an Owner opt-in.

 ### Requirements

@@ -429,7 +429,9 @@ claude --enable-auto-mode

 > **v2.1.112 update**: Auto mode no longer requires the `--enable-auto-mode` flag. Max subscribers access it directly on Opus 4.7.

-> **v2.1.158 update**: Auto mode is now available on Bedrock, Vertex, and Foundry for Opus 4.7/4.8 — **opt in** by setting `CLAUDE_CODE_ENABLE_AUTO_MODE=1`.
+> **v2.1.158 update**: Auto mode became available on Bedrock, Vertex, and Foundry for Opus 4.7/4.8, gated behind `CLAUDE_CODE_ENABLE_AUTO_MODE=1`.
+>
+> **v2.1.207 update**: That opt-in requirement was removed. Auto mode is now available by default on Bedrock, Vertex AI, Microsoft Foundry, and signed-in Claude apps gateway sessions for Claude Sonnet 5, Opus 4.7, Opus 4.8, and Fable 5 — no flag or env var needed. Administrators can disable it with `disableAutoMode` in managed settings. `CLAUDE_CODE_ENABLE_AUTO_MODE` is still accepted for compatibility but has no effect from v2.1.207 onward.

 Or set it as the default permission mode:

@@ -484,6 +486,12 @@ Auto mode blocks the following by default:
 claude auto-mode defaults
 ```

+**Restore default auto-mode configuration** (v2.1.212), with a confirmation prompt (`--yes` to skip):
+```bash
+claude auto-mode reset
+claude auto-mode reset --yes
+```
+
 **Configure trusted infrastructure** via the `autoMode.environment` managed setting for enterprise deployments. This allows administrators to define trusted CI/CD environments, deployment targets, and infrastructure patterns.

 #### Extending defaults with `"$defaults"` (v2.1.118)
@@ -872,7 +880,7 @@ Permission modes control what actions Claude can take without explicit approval.
 | `manual` | Read files only; prompts for all other actions. Renamed from `default` in v2.1.200 — `default` is still accepted as an alias |
 | `acceptEdits` | Read and edit files; prompts for commands |
 | `plan` | Read files only (research mode, no edits) |
-| `auto` | All actions with background safety classifier checks (Research Preview) |
+| `auto` | All actions with background safety classifier checks. Requires an eligible plan, model (Sonnet 5, Opus 4.7/4.8, or Fable 5 on most providers), and provider — see [Auto Mode](#auto-mode) |
 | `bypassPermissions` | All actions, no permission checks (dangerous) |
 | `dontAsk` | Only pre-approved tools execute; all others denied |

@@ -1110,7 +1118,8 @@ Manage multiple Claude Code sessions effectively.
 |---------|-------------|
 | `/resume` | Resume a conversation by ID or name |
 | `/rename` | Name the current session |
-| `/fork` | Fork current session into a new branch |
+| `/fork <directive>` | Spawn a background subagent that inherits the full conversation and works on the directive while you keep going |
+| `/branch [name]` | Switch into a copy of the conversation at this point, preserving the original |
 | `claude -c` | Continue most recent conversation |
 | `claude -r "session"` | Resume session by name or ID |

@@ -1131,15 +1140,25 @@ claude -r "auth-refactor" "finish this PR"
 /rename auth-refactor
 ```

-### Forking Sessions
+> **v2.1.212 update**: Typing `/resume` (with no arguments) in the agent view now opens a picker of past sessions — including sessions removed from the visible list — and resumes the chosen one as a background session.

-Fork a session to try an alternative approach without losing the original:
+### Forking and Branching Sessions
+
+`/fork <directive>` spawns a background subagent that inherits the full conversation and works on the directive while you keep working in the current session — its own row in `claude agents`, with the result returned to your conversation when it finishes:

 ```
-/fork
+/fork Investigate why the auth tests are flaky
 ```

-Or from the CLI:
+To switch into a copy of the conversation yourself instead of delegating to a background subagent, use `/branch [name]`, which preserves the original and lets you return to it with `/resume`:
+
+```
+/branch try-oauth-instead
+```
+
+> **Note**: Before v2.1.161, `/fork` was simply an alias for `/branch`. Since v2.1.161 they are distinct: `/fork` delegates to a background subagent, `/branch` switches you into a copy in place.
+
+Or fork from the CLI:
 ```bash
 claude --resume auth-refactor --fork-session "try OAuth instead"
 ```
@@ -1216,6 +1235,24 @@ Claude Code supports keyboard shortcuts for efficiency. Here's the complete refe
 | `Tab` | Autocomplete |
 | `↑ / ↓` | Command history |

+### Accessibility
+
+Screen reader mode (v2.1.208+) switches the CLI to a plain-text rendering mode designed for screen readers. Enable it with the CLI flag, an environment variable, or a settings key:
+
+```bash
+claude --ax-screen-reader
+```
+
+```bash
+export CLAUDE_AX_SCREEN_READER=1
+```
+
+```json
+{
+  "axScreenReader": true
+}
+```
+
 ### Customizing keybindings

 Create custom keyboard shortcuts by running `/keybindings`, which opens `~/.claude/keybindings.json` for editing (v2.1.18+).
@@ -2210,6 +2247,12 @@ export ANTHROPIC_WORKSPACE_ID=ws_abc123                     # Scope the federate

 # Stop hook safety cap (v2.1.143+)
 export CLAUDE_CODE_STOP_HOOK_BLOCK_CAP=8                    # Max consecutive Stop-hook blocks before the session ends with a warning. Set 0 to disable the cap.
+
+# Session-wide spawn caps (v2.1.212)
+export CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION=200         # Cap on WebSearch tool calls per session, to stop runaway search loops. Default 200.
+
+# Accessibility (v2.1.208)
+export CLAUDE_AX_SCREEN_READER=1                            # Enable plain-text screen reader rendering mode. Same effect as --ax-screen-reader or "axScreenReader": true in settings.
 ```

 > **v2.1.108**: `ENABLE_PROMPT_CACHING_1H=1` — use a 1-hour prompt cache TTL instead of the default 5-minute TTL. Reduces cache misses in long, stable sessions. (v2.1.129 fixes a regression where the 1-hour TTL was silently downgraded to 5 minutes.)
@@ -2382,8 +2425,8 @@ For more information about Claude Code and related features:

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
 - https://docs.anthropic.com/en/docs/claude-code/settings
@@ -2401,4 +2444,5 @@ For more information about Claude Code and related features:
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.154
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/sub-agents
+- https://code.claude.com/docs/en/commands
 **Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -533,9 +533,9 @@ Proceed with this plan? (yes/no/modify)
 7. **Plan for testing** at each phase

 ---
-**Last Updated**: June 2, 2026
-**Claude Code Version**: 2.1.160
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/changelog
 - https://code.claude.com/docs/en/best-practices
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -53,6 +53,7 @@ The older JavaScript bundle is still produced for Windows and for environments t
 | `claude mcp serve` | Run Claude Code as an MCP server | `claude mcp serve` |
 | `claude agents` | Open the **Agent View** (Research Preview, v2.1.139+) — multi-session manager listing every Claude Code session with its status. See [Agent View](#agent-view-claude-agents-v21139) below. | `claude agents` |
 | `claude auto-mode defaults` | Print auto mode default rules as JSON | `claude auto-mode defaults` |
+| `claude auto-mode reset` | Restore default auto-mode configuration, with a confirmation prompt (`--yes` to skip) (v2.1.212) | `claude auto-mode reset --yes` |
 | `claude remote-control` | Start Remote Control server | `claude remote-control` |
 | `claude plugin` | Manage plugins (install, enable, disable) | `claude plugin install my-plugin` |
 | `claude plugin init <name>` | Scaffold a new plugin in `.claude/skills` — auto-loads with no marketplace required (v2.1.157+) | `claude plugin init my-plugin` |
@@ -369,6 +370,7 @@ claude project purge --all --interactive
 | `--fork-session` | Create new session ID when resuming | `claude --resume abc --fork-session` |
 | `--max-budget-usd` | Maximum spend (print mode) | `claude -p --max-budget-usd 5.00 "query"` |
 | `--json-schema` | Validated JSON output | `claude -p --json-schema '{"type":"object"}' "q"` |
+| `--ax-screen-reader` | Plain-text rendering mode for screen readers (v2.1.208) | `claude --ax-screen-reader` |

 ### Platform & Theme Notes (v2.1.112)

@@ -845,7 +847,11 @@ The "ultrathink" keyword in prompts activates deep reasoning. The `/effort` menu
 | `CLAUDE_CODE_FORCE_SYNC_OUTPUT` | Set to `1` to force synchronous output for terminals where auto-detection misses (e.g., Emacs `eat`) (v2.1.129+) |
 | `CLAUDE_CODE_PACKAGE_MANAGER_AUTO_UPDATE` | Set to `1` to enable background upgrades for Homebrew/WinGet installs (which normally do not auto-update) (v2.1.129+) |
 | `CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY` | Set to `1` to opt in to gateway `/v1/models` discovery when `ANTHROPIC_BASE_URL` is set. Without it, `/model` shows the built-in static list (v2.1.129+) |
-| `CLAUDE_CODE_ENABLE_AUTO_MODE` | Set to `1` to opt in to auto mode on Bedrock, Vertex, and Foundry for Opus 4.7/4.8 (v2.1.158+) |
+| `CLAUDE_CODE_ENABLE_AUTO_MODE` | Legacy opt-in for auto mode on Bedrock, Vertex, and Foundry (v2.1.158–v2.1.206). As of v2.1.207, auto mode is available by default on those providers for Sonnet 5, Opus 4.7/4.8, and Fable 5 — this variable is accepted for compatibility but has no effect |
+| `CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION` | Cap on WebSearch tool calls per session, to stop runaway search loops. Default 200 (v2.1.212) |
+| `CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION` | Cap on subagent spawns per session, to stop runaway delegation loops. Default 200; `/clear` resets the budget (v2.1.212) |
+| `CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS` | Threshold, in milliseconds, before a long-running MCP tool call auto-backgrounds. Default 120000 (2 minutes) (v2.1.212) |
+| `CLAUDE_AX_SCREEN_READER` | Set to `1` to enable plain-text screen reader rendering mode. Same effect as `--ax-screen-reader` or `"axScreenReader": true` in settings (v2.1.208) |
 | `CLAUDE_CLIENT_PRESENCE_FILE` | Point at a marker file to suppress mobile push notifications while you're at the machine (v2.1.181+). Note: the name is `CLAUDE_CLIENT_PRESENCE_FILE`, not `CLAUDE_CODE_CLIENT_PRESENCE_FILE`. |
 | `CLAUDE_CODE_MAX_RETRIES` | Maximum number of API retry attempts. Capped at 15 as of v2.1.186. |
 | `CLAUDE_CODE_RETRY_WATCHDOG` | Retry control recommended for unattended sessions, as an alternative to raising `CLAUDE_CODE_MAX_RETRIES` (v2.1.186+). |
@@ -980,8 +986,8 @@ claude -p --output-format json "query"

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/cli-reference
 - https://code.claude.com/docs/en/env-vars
@@ -17,12 +17,12 @@
 |---------|----------|----------|-------|-----------|
 | **Slash Commands** | 60+ | 8 | 68+ | [01-slash-commands/](01-slash-commands/) |
 | **Subagents** | 6 | 11 | 17 | [04-subagents/](04-subagents/) |
-| **Skills** | 9 bundled | 6 | 15 | [03-skills/](03-skills/) |
+| **Skills** | 10 bundled | 6 | 16 | [03-skills/](03-skills/) |
 | **Plugins** | - | 3 | 3 | [07-plugins/](07-plugins/) |
 | **MCP Servers** | 1 | 8 | 9 | [05-mcp/](05-mcp/) |
 | **Hooks** | 29 events | 8 | 8 | [06-hooks/](06-hooks/) |
 | **Memory** | 7 types | 3 | 3 | [02-memory/](02-memory/) |
-| **Total** | **103** | **47** | **125** | |
+| **Total** | **104** | **47** | **126** | |

 ---

@@ -130,11 +130,11 @@ Claude Code supports 6 permission modes that control how tool use is authorized.
 | `manual` | Prompt for each tool call | Standard interactive use (renamed from `default` in v2.1.200; `default` still accepted) |
 | `acceptEdits` | Auto-accept file edits, prompt for others | Trusted editing workflows |
 | `plan` | Read-only tools only, no writes | Planning and exploration |
-| `auto` | Accept all tools without prompting | Fully autonomous operation (Research Preview) |
+| `auto` | All actions, with background safety classifier checks | Long tasks, reducing prompt fatigue |
 | `bypassPermissions` | Skip all permission checks | CI/CD, headless environments |
 | `dontAsk` | Skip tools that would require permission | Non-interactive scripting |

-> **Note**: `auto` mode is a Research Preview feature (March 2026). Use `bypassPermissions` only in trusted, sandboxed environments.
+> **Note**: `auto` mode requires an eligible plan, model, and provider — see [09-advanced-features/](09-advanced-features/#auto-mode). Use `bypassPermissions` only in trusted, sandboxed environments.

 **Reference**: [Official Docs](https://code.claude.com/docs/en/permissions)

@@ -476,7 +476,7 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **Scheduled Tasks** | Set up recurring tasks with `/loop` and cron tools | Use `/loop 5m /command` or CronCreate tool |
 | **Chrome Integration** | Browser automation with headless Chromium | Use `--chrome` flag or `/chrome` command |
 | **Keyboard Customization** | Customize keybindings including chord support | Use `/keybindings` or edit `~/.claude/keybindings.json` |
-| **Auto Mode** | Fully autonomous operation without permission prompts (Research Preview) | Use `--mode auto` or `/permissions auto`; March 2026 |
+| **Auto Mode** | Fully autonomous operation with background safety classifier checks | `Shift+Tab` to cycle to it, or `--permission-mode auto` |
 | **Channels** | Multi-channel communication (Telegram, Slack, etc.) (Research Preview) | Configure channel plugins; March 2026 |
 | **Voice Dictation** | Voice input for prompts | Use microphone icon or voice keybinding |
 | **Agent Hook Type** | Hooks that spawn a subagent instead of running a shell command | Set `"type": "agent"` in hook configuration |
@@ -545,8 +545,8 @@ chmod +x ~/.claude/hooks/*.sh

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/commands
@@ -1,5 +1,101 @@
 # Changelog

+## [v2.1.212] — 2026-07-18
+
+### Synced to Claude Code v2.1.212
+
+Bumps tutorial coverage from the v2.1.206 baseline (2026-07-11 sync) to
+v2.1.212, plus a repo-internal accuracy audit that surfaced defects
+independent of the version delta.
+
+### Fixed
+
+- **Dead `#` memory-shortcut removed** — `02-memory/README.md` documented a
+  `#`-prefix quick-memory-add pattern in two places ("Method 3" and an
+  "Example 4" walkthrough) as a working feature, directly contradicting the
+  file's own commands table three lines from the top, which already marked
+  `#` **Discontinued**. Both sections were removed/rewritten to point at
+  `/memory` and conversational memory requests instead.
+- **Auto mode opt-in → opt-out on Bedrock/Vertex/Foundry (v2.1.207)** — auto
+  mode is now available by default on these providers (plus signed-in Claude
+  apps gateway sessions) for Sonnet 5, Opus 4.7/4.8, and Fable 5; the
+  `CLAUDE_CODE_ENABLE_AUTO_MODE` opt-in flag has no effect from v2.1.207
+  onward. Fixed in `09-advanced-features/README.md` and `10-cli/README.md`.
+- **`auto` permission mode mislabeled "Research Preview"** — current official
+  docs present `auto` as GA (available on all plans, gated only by model and
+  provider eligibility), not a preview feature. Corrected across
+  `09-advanced-features/README.md`, `CATALOG.md`, and `README.md`, which also
+  gained the missing `auto` row in its permission-mode summary.
+- **`/fork` / `/branch` history corrected** — the repo claimed "`/fork`
+  renamed to `/branch` in v2.1.77 (alias retained)." Current docs show these
+  are distinct, current commands: `/fork` spawns a background subagent that
+  inherits the conversation, `/branch` switches you into a copy in place.
+  They were a single aliased command only from v2.1.77 to v2.1.161. Fixed in
+  `01-slash-commands/README.md` and `09-advanced-features/README.md`.
+- **`effort` frontmatter enum completed** — `04-subagents/README.md` listed
+  4 levels (`low`/`medium`/`high`/`max`), missing `xhigh`, present in every
+  other file that documents this enum.
+- **Bundled-skills count reconciled** — `CATALOG.md`'s summary table said "9
+  bundled" against its own 10-row detail table; corrected to 10 (16 total).
+- **`INDEX.md` Feature Coverage Matrix fully recomputed** — header claimed "16
+  files," the Skills row claimed "28" for the same category; neither matched
+  the actual 21 files across the six documented skills. The Plugins row
+  similarly summed to 36 but showed 40. Every row was recounted by hand; the
+  Skills row now reads `5 | 9 | 7` (**21**) and the Plugins row now reads
+  `11 | 9 | 3 | 3 | 3 | 3 | 7` (**39**) — every row's Total now equals the sum
+  of its own cells.
+- **Stale `2.1.160` footer cluster bumped** — five files
+  (`08-checkpoints/checkpoint-examples.md`,
+  `09-advanced-features/planning-mode-examples.md`, and the three
+  `07-plugins/*/README.md` example-package files) still carried June 2026
+  footers with no version-specific content drift found; bumped to 2.1.212.
+- **`QUICK_REFERENCE.md` fully resynced** — was 52 versions behind at
+  `2.1.160`/June 2. Permission-mode block updated to `manual`
+  (formerly `default`), Compatible Models gained Sonnet 5, and the "New
+  Features (May 2026)" section was retitled and refreshed.
+- **`02-memory/README.md` Memory Hierarchy section rewritten** — replaced a
+  fabricated 8-tier strict-precedence model with the verified structure:
+  CLAUDE.md files and rules are concatenated into context (not selected by
+  override), and `managed-settings.d/` is a `settings.json` mechanism, not a
+  CLAUDE.md one. The Memory Architecture diagram was also corrected — it
+  previously conflated claude.ai's 24h synthesis cycle with Claude Code's
+  continuous auto memory.
+
+### Added
+
+- **Subagent output scanning (v2.1.210)** — Claude Code scans subagent
+  reports for prompt-injection patterns (fake system tags, fabricated
+  conversation turns, permission-bypass mentions) and neutralizes them.
+  Documented in a new subsection of `04-subagents/README.md`.
+- **Session-wide spawn caps (v2.1.212)** — default 200/session limits on
+  WebSearch calls (`CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION`) and subagent
+  spawns (`CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION`, reset by `/clear`).
+  Documented in `09-advanced-features/README.md`, `04-subagents/README.md`,
+  and `10-cli/README.md`.
+- **MCP long-running-tool auto-backgrounding (v2.1.212)** — MCP tool calls
+  over 2 minutes now auto-background instead of blocking the session;
+  configurable via `CLAUDE_CODE_MCP_AUTO_BACKGROUND_MS`. Documented in
+  `05-mcp/README.md`.
+- **`claude auto-mode reset` and `/resume` picker (v2.1.212)** — a new CLI
+  subcommand restores default auto-mode config; bare `/resume` now opens a
+  picker of past sessions (including removed ones) and resumes as a
+  background session. Documented in `09-advanced-features/README.md`,
+  `01-slash-commands/README.md`, and `10-cli/README.md`.
+- **Screen reader mode (v2.1.208)** — opt-in plain-text rendering via
+  `--ax-screen-reader`, `CLAUDE_AX_SCREEN_READER=1`, or `"axScreenReader":
+  true`. Documented in `09-advanced-features/README.md` and
+  `10-cli/README.md`.
+- **Task tool `mode` parameter deprecation (v2.1.212)** — noted in
+  `04-subagents/README.md`: subagents now inherit the parent session's
+  permission mode by default; the Task tool's per-call `mode` parameter is
+  ignored.
+
+### Known gaps (deferred, not fixed this sync)
+
+- `03-skills/.claude/skills/blog-draft/` is gitignored local testing scratch
+  (`# Local skill testing` in `03-skills/.gitignore`), not a tracked
+  duplicate of `03-skills/blog-draft/` — confirmed no action needed.
+
 ## [v2.1.160] — 2026-06-02

 ### Synced to Claude Code v2.1.160
@@ -60,7 +60,7 @@ Persistent context and project standards.

 ---

-## 03. Skills (16 files)
+## 03. Skills (21 files)

 Auto-invoked capabilities with scripts and templates.

@@ -767,11 +767,11 @@ Run tests in background
 |----------|----------|--------|-----|-------|---------|-----------|------|--------|-------|
 | **01 Slash Commands** | 8 | - | - | - | - | - | 1 | 1 | **10** |
 | **02 Memory** | - | - | - | - | - | 3 | 1 | 2 | **6** |
-| **03 Skills** | - | - | - | - | 5 | 9 | 1 | - | **28** |
+| **03 Skills** | - | - | - | - | 5 | 9 | 7 | - | **21** |
 | **04 Subagents** | - | 8 | - | - | - | - | 1 | - | **9** |
 | **05 MCP** | - | - | 4 | - | - | - | 1 | - | **5** |
 | **06 Hooks** | - | - | - | 8 | - | - | 1 | - | **9** |
-| **07 Plugins** | 11 | 9 | 3 | 3 | 3 | 3 | 4 | - | **40** |
+| **07 Plugins** | 11 | 9 | 3 | 3 | 3 | 3 | 7 | - | **39** |
 | **08 Checkpoints** | - | - | - | - | - | - | 1 | 1 | **2** |
 | **09 Advanced** | - | - | - | - | - | - | 1 | 2 | **3** |
 | **10 CLI** | - | - | - | - | - | - | 1 | - | **1** |
@@ -880,8 +880,8 @@ Want to add more examples? Follow the structure:

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/hooks
@@ -92,17 +92,18 @@ chmod +x ~/.claude/hooks/*.sh
 /plan Task description

 # Permission modes (use --permission-mode flag)
-# default        - Ask for approval on risky actions
+# manual         - Ask for approval on every action (formerly "default"; "default" still accepted as alias)
 # acceptEdits    - Auto-accept file edits, ask for others
 # plan           - Read-only analysis, no modifications
-# dontAsk        - Accept all actions except risky ones
 # auto           - Background classifier decides permissions automatically
+# dontAsk        - Only pre-approved tools run; everything else is denied
 # bypassPermissions - Accept all actions (requires --dangerously-skip-permissions)

 # Session management
-/resume                # Resume a previous conversation
+/resume                # Resume a previous conversation (no args = picker of past sessions)
 /rename "name"         # Name the current session
-/fork                  # Fork the current session
+/fork <directive>      # Spawn a background subagent that inherits the conversation
+/branch [name]         # Switch into a copy of the conversation, preserving the original
 claude -c              # Continue most recent conversation
 claude -r "session"    # Resume session by name/ID
 ```
@@ -382,11 +383,11 @@ cp -r 03-skills/code-review-specialist ~/.claude/skills/

 ---

-## New Features (May 2026)
+## Feature Highlights

 | Feature | Description | Usage |
 |---------|-------------|-------|
-| **Auto Mode** | Fully autonomous operation with background classifier | `--enable-auto-mode` flag, `Shift+Tab` to cycle modes |
+| **Auto Mode** | Fully autonomous operation with background classifier; available by default on Bedrock/Vertex/Foundry as of v2.1.207 | `Shift+Tab` to cycle modes, or `--permission-mode auto` |
 | **Channels** | Discord and Telegram integration | `--channels` flag, Discord/Telegram bots |
 | **Voice Dictation** | Speak commands and context to Claude | `/voice` command |
 | **Hooks (29 events)** | Expanded hook system with 5 types | command, http, mcp_tool, prompt, agent hook types |
@@ -408,6 +409,9 @@ cp -r 03-skills/code-review-specialist ~/.claude/skills/
 | **/run** *(v2.1.145+)* | Launch this project's app to see a change running | `/run` |
 | **/verify** *(v2.1.145+)* | Build, run, and observe the app to confirm a fix works | `/verify` |
 | **/run-skill-generator** *(v2.1.145+)* | Teach `/run`/`/verify` how to handle a specific project | `/run-skill-generator` |
+| **Subagent Output Scanning** *(v2.1.210+)* | Scans subagent reports for prompt-injection patterns and neutralizes them | On by default, no opt-out |
+| **Session-Wide Spawn Caps** *(v2.1.212)* | 200/session limits on WebSearch calls and subagent spawns, to stop runaway loops | `CLAUDE_CODE_MAX_WEB_SEARCHES_PER_SESSION`, `CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION`; `/clear` resets |
+| **Screen Reader Mode** *(v2.1.208)* | Plain-text rendering mode for screen readers | `--ax-screen-reader` flag, `CLAUDE_AX_SCREEN_READER=1`, or `"axScreenReader": true` in settings |

 ---

@@ -509,12 +513,14 @@ Getting started checklist:
 **This Card**: Keep it handy for quick reference!

 ---
-**Last Updated**: June 2, 2026
-**Claude Code Version**: 2.1.160
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/hooks
 - https://code.claude.com/docs/en/commands
+- https://code.claude.com/docs/en/permission-modes
+- https://github.com/anthropics/claude-code/blob/main/CHANGELOG.md
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.153
 - https://github.com/anthropics/claude-code/releases/tag/v2.1.154
-**Compatible Models**: Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
+**Compatible Models**: Claude Sonnet 5, Claude Sonnet 4.6, Claude Opus 4.8, Claude Haiku 4.5
@@ -12,7 +12,7 @@
 [![GitHub Stars](https://img.shields.io/github/stars/luongnv89/claude-howto?style=flat&color=gold)](https://github.com/luongnv89/claude-howto/stargazers)
 [![GitHub Forks](https://img.shields.io/github/forks/luongnv89/claude-howto?style=flat)](https://github.com/luongnv89/claude-howto/network/members)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](LICENSE)
-[![Version](https://img.shields.io/badge/version-2.1.206-brightgreen)](CHANGELOG.md)
+[![Version](https://img.shields.io/badge/version-2.1.212-brightgreen)](CHANGELOG.md)
 [![Claude Code](https://img.shields.io/badge/Claude_Code-2.1+-purple)](https://code.claude.com)

 🌐 **Language / Ngôn ngữ / 语言 / Мова:** [English](README.md) | [Tiếng Việt](vi/README.md) | [中文](zh/README.md) | [Українська](uk/README.md) | [日本語](ja/README.md)
@@ -102,7 +102,7 @@ Run `/lesson-quiz [topic]` after each module. The quiz pinpoints what you missed

 - **GitHub stars** from developers who use Claude Code daily
 - **Forks** from teams adapting this guide for their own workflows
- **Actively maintained** — synced with every Claude Code release (latest: v2.1.206, July 2026)
+- **Actively maintained** — synced with every Claude Code release (latest: v2.1.212, July 2026)
 - **Community-driven** — contributions from developers who share their real-world configurations

 [![Star History Chart](https://api.star-history.com/svg?repos=luongnv89/claude-howto&type=Date)](https://star-history.com/#luongnv89/claude-howto&Date)
@@ -201,7 +201,7 @@ cp -r 03-skills/code-review-specialist ~/.claude/skills/
 Yes. MIT licensed, free forever. Use it in personal projects, at work, in your team — no restrictions beyond including the license notice.

 **Is this maintained?**
-Actively. The guide is synced with every Claude Code release. Current version: v2.1.206 (July 2026), compatible with Claude Code 2.1+.
+Actively. The guide is synced with every Claude Code release. Current version: v2.1.212 (July 2026), compatible with Claude Code 2.1+.

 **How is this different from the official docs?**
 The official docs are a feature reference. This guide is a tutorial with diagrams, production-ready templates, and a progressive learning path. They complement each other — start here to learn, reference the docs when you need specifics.
@@ -559,9 +559,9 @@ Configure hooks in `~/.claude/settings.json`:
 - **Planning Mode** — Create detailed implementation plans before coding
 - **Extended Thinking** — Deep reasoning for complex problems (toggle with `Alt+T` / `Option+T`)
 - **Background Tasks** — Run long operations without blocking
- **Permission Modes** — `manual` (formerly `default`; `default` still accepted), `acceptEdits`, `plan`, `dontAsk`, `bypassPermissions`
+- **Permission Modes** — `manual` (formerly `default`; `default` still accepted), `acceptEdits`, `plan`, `auto`, `dontAsk`, `bypassPermissions`
 - **Headless Mode** — Run Claude Code in CI/CD: `claude -p "Run tests and generate report"`
- **Session Management** — `/resume`, `/rename`, `/fork`, `claude -c`, `claude -r`
+- **Session Management** — `/resume`, `/rename`, `/fork`, `/branch`, `claude -c`, `claude -r`
 - **Configuration** — Customize behavior in `~/.claude/settings.json`

 See [config-examples.json](09-advanced-features/config-examples.json) for complete configurations.
@@ -871,8 +871,8 @@ MIT License - see [LICENSE](LICENSE). Free to use, modify, and distribute. The o

 ---

-**Last Updated**: July 11, 2026
-**Claude Code Version**: 2.1.206
+**Last Updated**: July 18, 2026
+**Claude Code Version**: 2.1.212
 **Sources**:
 - https://code.claude.com/docs/en/overview
 - https://code.claude.com/docs/en/changelog
@@ -51,7 +51,7 @@ Claude Code の高度な機能は、プランニング、推論、自動化、
 **主な高度な機能：**
 - **プランニングモード**：コーディング前に詳細な実装計画を作成する
 - **拡張思考**：複雑な問題に対する深い推論を行う
- **オートモード**：バックグラウンドの安全分類器が各アクションを実行前にレビューする（Research Preview）
+- **オートモード**：バックグラウンドの安全分類器が各アクションを実行前にレビューする
 - **バックグラウンドタスク**：会話をブロックせずに長時間処理を実行する
 - **権限モード**：Claude が実行可能な操作を制御する（`default`、`acceptEdits`、`plan`、`auto`、`dontAsk`、`bypassPermissions`）
 - **プリントモード**：自動化や CI/CD のために Claude Code を非対話で実行する（`claude -p`）
@@ -408,7 +408,7 @@ export CLAUDE_CODE_EFFORT_LEVEL=xhigh

 ## オートモード

-オートモードは Research Preview の権限モード（2026 年 3 月）で、バックグラウンドの安全分類器が各アクションを実行前にレビューする。Claude が自律的に作業しつつ、危険な操作はブロックできる。
+オートモードは、バックグラウンドの安全分類器が各アクションを実行前にレビューする権限モードである。Claude が自律的に作業しつつ、危険な操作はブロックできる。すべてのプランで利用可能だが、対象モデルおよびプロバイダの条件を満たす必要がある。

 ### 必要要件

@@ -805,7 +805,7 @@ export CLAUDE_CODE_DISABLE_CRON=1
 | `default` | ファイル読み取りのみ。それ以外の操作は確認を求める |
 | `acceptEdits` | ファイル読み取りと編集を許可。コマンドは確認を求める |
 | `plan` | ファイル読み取りのみ（リサーチモード、編集なし） |
-| `auto` | バックグラウンドの安全分類器のチェック付きで全操作を許可（Research Preview） |
+| `auto` | バックグラウンドの安全分類器のチェック付きで全操作を許可 |
 | `bypassPermissions` | 全操作、権限チェックなし（危険） |
 | `dontAsk` | 事前承認済みツールのみ実行、その他はすべて拒否 |

@@ -132,11 +132,11 @@ Claude Code はツール使用の許可を制御する 6 つの権限モード
 | `default` | ツール呼び出しごとに確認 | 標準的な対話利用 |
 | `acceptEdits` | ファイル編集を自動承認、それ以外は確認 | 信頼できる編集ワークフロー |
 | `plan` | 読み取り専用ツールのみ、書き込みなし | 計画と探索 |
-| `auto` | プロンプトなしですべてのツールを承認 | 完全自律動作（Research Preview） |
+| `auto` | バックグラウンドの安全分類器のチェック付きで全操作を許可 | 長時間タスク、確認プロンプトの削減 |
 | `bypassPermissions` | すべての権限チェックをスキップ | CI/CD、ヘッドレス環境 |
 | `dontAsk` | 権限が必要なツールをスキップ | 非対話スクリプト |

-> **注**：`auto` モードは Research Preview 機能（2026 年 3 月）。`bypassPermissions` は信頼されたサンドボックス環境でのみ利用すること。
+> **注**：`auto` モードは対象プラン・モデル・プロバイダの条件を満たす必要がある — [09-advanced-features/](../09-advanced-features/#auto-mode) を参照。`bypassPermissions` は信頼されたサンドボックス環境でのみ利用すること。

 **リファレンス**：[公式ドキュメント](https://code.claude.com/docs/en/permissions)

@@ -467,7 +467,7 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **スケジュールタスク** | `/loop` および cron ツールで反復タスクを設定 | `/loop 5m /command` または CronCreate ツールを使う |
 | **Chrome 連携** | ヘッドレス Chromium によるブラウザ自動化 | `--chrome` フラグまたは `/chrome` コマンドを使う |
 | **キーバインドカスタマイズ** | コード対応を含むキーバインドのカスタマイズ | `/keybindings` を使うか `~/.claude/keybindings.json` を編集 |
-| **Auto Mode** | 権限プロンプトなしの完全自律動作（Research Preview） | `--mode auto` または `/permissions auto` を使う、2026 年 3 月 |
+| **Auto Mode** | バックグラウンドの安全分類器チェック付きで完全自律動作 | `Shift+Tab` で切り替え、または `--permission-mode auto` |
 | **チャンネル** | 複数チャネル通信（Telegram、Slack など）（Research Preview） | チャンネルプラグインを設定、2026 年 3 月 |
 | **音声入力** | プロンプトの音声入力 | マイクアイコンまたは音声キーバインドを使う |
 | **Agent フック種別** | シェルコマンド実行ではなくサブエージェントを生成するフック | フック設定で `"type": "agent"` を指定 |
@@ -44,7 +44,7 @@
 **Ключові розширені функції:**
 - **Режим планування**: Створення детальних планів реалізації перед кодуванням
 - **Розширене мислення**: Глибоке міркування для складних проблем
- **Auto Mode**: Фоновий класифікатор безпеки перевіряє кожну дію перед виконанням (Research Preview)
+- **Auto Mode**: Фоновий класифікатор безпеки перевіряє кожну дію перед виконанням
 - **Фонові завдання**: Виконання тривалих операцій без блокування розмови
 - **Режими дозволів**: Контроль дій Claude (`default`, `acceptEdits`, `plan`, `auto`, `dontAsk`, `bypassPermissions`)
 - **Режим друку**: Неінтерактивний запуск Claude Code для автоматизації та CI/CD (`claude -p`)
@@ -348,7 +348,7 @@ export CLAUDE_CODE_EFFORT_LEVEL=high

 ## Auto Mode

-Auto Mode — це режим дозволів Research Preview (березень 2026), який використовує фоновий класифікатор безпеки для перевірки кожної дії перед виконанням. Він дозволяє Claude працювати автономно, блокуючи небезпечні операції.
+Auto Mode — це режим дозволів, який використовує фоновий класифікатор безпеки для перевірки кожної дії перед виконанням. Він дозволяє Claude працювати автономно, блокуючи небезпечні операції. Доступний на всіх планах, але вимагає відповідної моделі та провайдера.

 ### Вимоги

@@ -678,7 +678,7 @@ export CLAUDE_CODE_DISABLE_CRON=1
 | `default` | Тільки читання файлів; запитує дозвіл на всі інші дії |
 | `acceptEdits` | Читання та редагування файлів; запитує дозвіл на команди |
 | `plan` | Тільки читання файлів (режим дослідження, без редагувань) |
-| `auto` | Усі дії з перевірками фонового класифікатора безпеки (Research Preview) |
+| `auto` | Усі дії з перевірками фонового класифікатора безпеки |
 | `bypassPermissions` | Усі дії, без перевірки дозволів (небезпечно) |
 | `dontAsk` | Виконуються лише попередньо затверджені інструменти; всі інші відхиляються |

@@ -124,11 +124,11 @@ Claude Code підтримує 6 режимів дозволів, що конт
 | `default` | Запитувати дозвіл на кожен виклик | Стандартна інтерактивна робота |
 | `acceptEdits` | Автоприйняття редагувань, запит на інше | Довірені процеси редагування |
 | `plan` | Лише інструменти читання, без запису | Планування та дослідження |
-| `auto` | Приймати всі інструменти без запиту | Повністю автономна робота (Research Preview) |
+| `auto` | Усі дії з перевірками фонового класифікатора безпеки | Тривалі завдання, менше запитів на дозвіл |
 | `bypassPermissions` | Пропустити всі перевірки дозволів | CI/CD, headless-середовища |
 | `dontAsk` | Пропустити інструменти, що потребують дозволу | Неінтерактивне скриптування |

-> **Примітка**: Режим `auto` — це Research Preview (березень 2026). Використовуйте `bypassPermissions` лише в довірених ізольованих середовищах.
+> **Примітка**: Режим `auto` вимагає відповідного плану, моделі та провайдера — див. [09-advanced-features/](../09-advanced-features/#auto-mode). Використовуйте `bypassPermissions` лише в довірених ізольованих середовищах.

 **Довідник**: [Офіційна документація](https://code.claude.com/docs/en/permissions)

@@ -457,7 +457,7 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **Заплановані завдання** | Повторювані завдання з `/loop` та cron | `/loop 5m /command` або CronCreate |
 | **Інтеграція з Chrome** | Автоматизація браузера через headless Chromium | Прапорець `--chrome` або `/chrome` |
 | **Кастомізація клавіатури** | Налаштування клавіш з підтримкою акордів | `/keybindings` або `~/.claude/keybindings.json` |
-| **Auto Mode** | Повністю автономна робота (Research Preview) | `--mode auto` або `/permissions auto` |
+| **Auto Mode** | Повністю автономна робота з перевірками фонового класифікатора безпеки | `Shift+Tab` для перемикання, або `--permission-mode auto` |
 | **Канали** | Багатоканальна комунікація (Telegram, Slack тощо) (Research Preview) | Налаштування канальних плагінів |
 | **Голосовий ввід** | Голосове введення промптів | Іконка мікрофона або клавіша |
 | **Тип хука Agent** | Хуки, що запускають субагента замість shell-команди | `"type": "agent"` у конфігурації хука |
@@ -43,7 +43,7 @@ Advanced features in Claude Code extend the core capabilities with planning, rea
 **Key advanced features include:**
 - **Planning Mode**: Create detailed implementation plans before coding
 - **Extended Thinking**: Deep reasoning for complex problems
- **Auto Mode**: Background safety classifier reviews each action before execution (Research Preview)
+- **Auto Mode**: Background safety classifier reviews each action before execution
 - **Background Tasks**: Run long operations without blocking the conversation
 - **Permission Modes**: Control what Claude can do (`default`, `acceptEdits`, `plan`, `auto`, `dontAsk`, `bypassPermissions`)
 - **Print Mode**: Run Claude Code non-interactively for automation and CI/CD (`claude -p`)
@@ -343,7 +343,7 @@ Toggle during a session with `Alt+T` / `Option+T`, set effort with `/effort`, or

 ## Auto Mode

-Auto Mode is a Research Preview permission mode (March 2026) that uses a background safety classifier to review each action before execution. It allows Claude to work autonomously while blocking dangerous operations.
+Auto Mode is a permission mode that uses a background safety classifier to review each action before execution. It allows Claude to work autonomously while blocking dangerous operations. It's available on all plans, but requires an eligible model and provider.

 ### Requirements

@@ -672,7 +672,7 @@ Permission modes control what actions Claude can take without explicit approval.
 | `default` | Read files only; prompts for all other actions |
 | `acceptEdits` | Read and edit files; prompts for commands |
 | `plan` | Read files only (research mode, no edits) |
-| `auto` | All actions with background safety classifier checks (Research Preview) |
+| `auto` | All actions with background safety classifier checks |
 | `bypassPermissions` | All actions, no permission checks (dangerous) |
 | `dontAsk` | Only pre-approved tools execute; all others denied |

@@ -120,11 +120,11 @@ Claude Code hỗ trợ 6 permission modes điều khiển cách tool use đượ
 | `default` | Hỏi cho mỗi tool call | Sử dụng tương tác tiêu chuẩn |
 | `acceptEdits` | Tự động chấp nhận edits file, hỏi cho các khác | Workflows chỉnh sửa tin cậy |
 | `plan` | Chỉ tools read-only, không ghi | Lập kế hoạch và khám phá |
-| `auto` | Chấp nhận tất cả tools mà không hỏi | Vận hành hoàn toàn tự động (Research Preview) |
+| `auto` | Tất cả hành động, có kiểm tra từ bộ phân loại an toàn nền | Tác vụ dài, giảm số lần nhắc quyền |
 | `bypassPermissions` | Bỏ qua tất cả kiểm tra permissions | CI/CD, môi trường headless |
 | `dontAsk` | Bỏ qua tools cần yêu cầu permission | Scripting không tương tác |

-> **Lưu Ý**: `auto` mode là tính năng Research Preview (Tháng 3 năm 2026). Chỉ dùng `bypassPermissions` trong môi trường tin cậy, được sandbox.
+> **Lưu Ý**: `auto` mode yêu cầu gói dịch vụ, model, và nhà cung cấp đủ điều kiện — xem [09-advanced-features/](../09-advanced-features/#auto-mode). Chỉ dùng `bypassPermissions` trong môi trường tin cậy, được sandbox.

 **Tham Chiếu**: [Tài Liệu Chính Thức](https://code.claude.com/docs/en/permissions)

@@ -445,7 +445,7 @@ cp ../02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **Scheduled Tasks** | Thiết lập tasks định kỳ với `/loop` và cron tools | Sử dụng `/loop 5m /command` hoặc công cụ CronCreate |
 | **Chrome Integration** | Tự động hóa trình duyệt với headless Chromium | Sử dụng flag `--chrome` hoặc command `/chrome` |
 | **Keyboard Customization** | Tùy chỉnh keybindings bao gồm hỗ trợ chord | Sử dụng `/keybindings` hoặc chỉnh sửa `~/.claude/keybindings.json` |
-| **Auto Mode** | Vận hành hoàn toàn tự động mà không cần prompts permission (Research Preview) | Sử dụng `--mode auto` hoặc `/permissions auto`; Tháng 3 năm 2026 |
+| **Auto Mode** | Vận hành hoàn toàn tự động với kiểm tra từ bộ phân loại an toàn nền | `Shift+Tab` để chuyển sang, hoặc `--permission-mode auto` |
 | **Channels** | Communication đa kênh (Telegram, Slack, v.v.) (Research Preview) | Cấu hình plugins channels; Tháng 3 năm 2026 |
 | **Voice Dictation** | Input giọng nói cho prompts | Sử dụng icon microphone hoặc keybinding giọng nói |
 | **Agent Hook Type** | Hooks tạo subagent thay vì chạy shell command | Đặt `"type": "agent"` trong cấu hình hook |
@@ -44,7 +44,7 @@ Claude Code 的高级功能把基础能力扩展到了规划、推理、自动
 **核心高级功能包括：**
 - **Planning Mode**：先写详细实现计划，再开始编码
 - **Extended Thinking**：对复杂问题进行更深入的推理
- **Auto Mode**：由后台安全分类器在每一步执行前进行审查（Research Preview）
+- **Auto Mode**：由后台安全分类器在每一步执行前进行审查
 - **Background Tasks**：长时间任务不阻塞对话
 - **Permission Modes**：控制 Claude 可以做什么（`default`、`acceptEdits`、`plan`、`auto`、`dontAsk`、`bypassPermissions`）
 - **Print Mode**：非交互式运行，适合自动化和 CI/CD（`claude -p`）
@@ -247,7 +247,7 @@ Recommendation
 <a id="auto-mode"></a>
 ## 自动模式

-Auto mode 是一个研究预览功能。它会在每个动作执行前，用后台安全分类器评估风险，帮助你在自动化和安全之间取得平衡。
+Auto mode 是一种权限模式，会在每个动作执行前用后台安全分类器评估风险，帮助你在自动化和安全之间取得平衡。所有套餐均可使用，但需要满足合格的模型和提供商条件。

 ### 需求

@@ -119,11 +119,11 @@ Claude Code 提供 6 种权限模式，用来控制工具调用如何被授权
 | `default` | 每次工具调用都询问 | 标准交互式使用 |
 | `acceptEdits` | 自动接受文件编辑，其他情况仍询问 | 可信编辑工作流 |
 | `plan` | 只允许只读工具，不允许写入 | 规划与探索 |
-| `auto` | 不再提示，自动接受所有工具 | 完全自主运行（Research Preview） |
+| `auto` | 所有操作，附带后台安全分类器检查 | 长任务、减少权限提示 |
 | `bypassPermissions` | 跳过所有权限检查 | CI/CD、无头环境 |
 | `dontAsk` | 跳过需要权限的工具 | 非交互脚本 |

-> **注意**：`auto` 模式是 Research Preview 功能（2026 年 3 月）。只有在可信且已隔离的环境中才使用 `bypassPermissions`。
+> **注意**：`auto` 模式需要满足合格的套餐、模型和提供商条件 — 参见 [09-advanced-features/](../09-advanced-features/#auto-mode)。只有在可信且已隔离的环境中才使用 `bypassPermissions`。

 **参考**: [官方文档](https://code.claude.com/docs/en/permissions)

@@ -444,7 +444,7 @@ cp 02-memory/personal-CLAUDE.md ~/.claude/CLAUDE.md
 | **Scheduled Tasks** | 使用 `/loop` 和 cron 工具设置周期任务 | 使用 `/loop 5m /command` 或 CronCreate 工具 |
 | **Chrome Integration** | 使用无头 Chromium 做浏览器自动化 | 使用 `--chrome` 标志或 `/chrome` 命令 |
 | **Keyboard Customization** | 自定义按键映射并支持 chord | 使用 `/keybindings` 或编辑 `~/.claude/keybindings.json` |
-| **自动模式（Auto Mode）** | 无需权限提示的完全自主运行（Research Preview） | 使用 `--mode auto` 或 `/permissions auto`；2026 年 3 月 |
+| **自动模式（Auto Mode）** | 附带后台安全分类器检查的完全自主运行 | `Shift+Tab` 切换，或使用 `--permission-mode auto` |
 | **通道（Channels）** | 多通道通信（Telegram、Slack 等）（Research Preview） | 配置 channel plugins；2026 年 3 月 |
 | **语音输入（Voice Dictation）** | 用语音输入提示词 | 使用麦克风图标或语音快捷键 |
 | **Agent Hook Type** | 触发 subagent 而不是执行 shell 命令的 hook | 在 hook 配置中设置 `"type": "agent"` |