Claude Code Ultimate Guide: Updates

Claude Code v2.1.272

Mon, 14 Sep 2026 00:00:00 GMT

Bug fixes and reliability improvements; upstream provides no further detail

Claude Code v2.1.271

Mon, 14 Sep 2026 00:00:00 GMT

Added per-command allowed_domains to Bash, PowerShell and Monitor in auto mode with sandboxing: the hosts a command needs are reviewed with it and opened for it alone
Added omitClaudeMd to agent frontmatter and --agents JSON, letting custom and plugin subagents run without user, project and local CLAUDE.md files; managed policy files still load
Added --accept-command to claude plugin install/update, fast mode in Remote sessions, and mouse support in the /config panel
Fixed an unreadable enterprise managed-mcp.json being ignored: it now keeps exclusive MCP control and warns at startup, alongside ~80 further fixes across MCP, resume, hooks and terminal rendering

Breaking changes:

Monitor watches always have a deadline (30 minutes maximum, 10 in single-prompt -p runs); the no-timeout persistent option is removed.
In auto mode, a skill's or slash command's inline ! shell commands follow default-mode permission rules instead of the classifier.
Default dynamic workflow size is small on Pro plans, and the medium size guideline drops from 15 to 10 agents.

Claude Code v2.1.270

Sat, 12 Sep 2026 00:00:00 GMT

Fixed read-only git commands in Bash unexpectedly asking for permission after a session had been running for a while (regression in v2.1.269)

[New Section] September Security Review: AgentSec 2.29.0

Sat, 12 Sep 2026 00:00:00 GMT

Review the latest coding-agent and MCP advisories, GitSpawn delivery conditions, and documented fixed versions. The AgentSec feed tracks 131 CVE/advisory records; the searchable guide catalogue retains 140 records. New intelligence does not add executable detector coverage.

Claude Code v2.1.269

Fri, 11 Sep 2026 00:00:00 GMT

Added claude plugin eval for plugin evaluation suites, with scores and JSON/HTML reports
Added /output-style [name] for listing and switching styles in terminal, Remote Control, cloud and headless sessions
Fixed prompt-cache invalidation after interrupted or output-limited turns and /goal runs stalling after errors or limits
Security fixes cover plugin archive permissions, Bash tee write destinations and negated permission rules crossing settings sources

Claude Code v2.1.268

Thu, 10 Sep 2026 00:00:00 GMT

Fixed HTTP 400 errors on third-party Anthropic-compatible endpoints introduced in v2.1.265
Added a 300-second WebFetch deadline; CLAUDE_CODE_WEBFETCH_DEADLINE_MS overrides it (0 disables it)
Security fixes cover symlinked-path permission rules, untrusted teammate definitions and secrets shown in MCP/plugin diagnostics
Improved prompt-cache stability on Bedrock, Vertex and Foundry; /plugin changes now apply when its menu closes

Breaking changes:

Plain WebFetch deny/ask rules no longer gate Artifact reads or updates; use Artifact or WebFetch(domain:claude.ai).

Claude Code v2.1.267

Wed, 09 Sep 2026 00:00:00 GMT

Added maxEffortLevel globally or under modelSettings to cap effort across providers while allowing lower user choices
Added --system-prompt-snapshot off to render the system prompt anew for each request during prompt iteration
Fixed prompt-cache and thinking losses from reconnecting tools, model switches and session/subagent resumes
Security fixes reject unreadable managed hook/channel allowlists and close a marketplace backslash containment bypass

[New Page] AI-Assisted Open Source Contributions

Wed, 09 Sep 2026 00:00:00 GMT

Prepare a reviewable contribution with project policy, reproduction, author understanding and verification evidence. Includes a contribution packet, alongside new review-capacity and comprehension worksheets informed by IFTTD field accounts.

[New Section] Tokenade and RTK: Scope, Licence and Benchmark Evidence

Wed, 09 Sep 2026 00:00:00 GMT

The token-optimization guide adds Tokenade with its proprietary licence, client-specific coverage and vendor-maintained THOL results. RTK guidance distinguishes command-output estimates from full-session cost.

[New Page] Agentic Pentesting: DarkMoon and Strix

Wed, 09 Sep 2026 00:00:00 GMT

Compare execution architecture, privacy boundaries, and exploit evidence. Includes a sourced evaluation and a reproducible synthetic masking check.

Claude Code v2.1.266

Tue, 08 Sep 2026 00:00:00 GMT

Fixed v2.1.265 forcing Cloud-gateway sign-in when CLAUDE_CODE_USE_GATEWAY was set with alternative authentication; no configuration change is needed

Claude Code v2.1.265

Tue, 08 Sep 2026 00:00:00 GMT

Added --plugin-dir support for a parent folder of plugins, including child plugins added or removed during the session
Added a 1 GB limit for tool results saved to disk, with truncation disclosed in the conversation preview
Fixed resumed subagent prompt-cache reuse and preserved interrupted tool calls after a process crash
Security fixes close plugin backslash containment bypasses and prevent nested-repository git clean filters running during status/diff probes

Breaking changes:

Managed forceLoginGatewayUrl now selects gateway authentication at startup; leftover claude.ai credentials or API keys are not used.

Claude Code v2.1.263

Sun, 06 Sep 2026 00:00:00 GMT

Bug fixes and reliability improvements; upstream provides no further detail

[New Section] Skill Ownership, Evaluation and Retirement

Sat, 05 Sep 2026 00:00:00 GMT

The guide now separates personal, project, vendor, and marketplace ownership; distinguishes adaptable skills from deterministic harness controls; and treats usage reports and evals as bounded evidence. The landing adds adoption checks for distributed skills and a safe retirement path built around /skill-doctor, representative comparisons, and archives outside discovery paths.

Claude Code v2.1.261

Fri, 04 Sep 2026 00:00:00 GMT

Added /skill-doctor to show which loaded skills go unused and what they cost in context
Added bashOutputMaxChars/taskOutputMaxChars settings to raise inline command/task output limits up to 128K characters
Fixed resuming a session losing hook output and other context around parallel tool calls

Claude Code v2.1.260

Thu, 03 Sep 2026 00:00:00 GMT

Added a /diff panel that opens beside the conversation in fullscreen mode and shows uncommitted changes as Claude edits
Added a likely cause for prompt-cache misses to /cost and the status line
Fixed Edit/Write/Read permission rules with parentheses in the path being dropped, which left "read-only" folders writable
Fixed one uncompilable file permission pattern breaking every file edit with an "Invalid regular expression" error

Claude Code v2.1.259

Wed, 02 Sep 2026 00:00:00 GMT

Added managedMcpServers managed setting: organizations can provide HTTP/SSE MCP servers to every user
Added --permission-prompts none for unattended headless hosts
Fixed concurrent sessions silently reverting each other's ~/.claude.json changes, resetting workspace trust and losing MCP/project state
Fixed Bash Read() deny rules not covering files given as option values or git diff/git grep file operands

Claude Code v2.1.258

Tue, 01 Sep 2026 00:00:00 GMT

Fixed Claude Code failing to launch on macOS 12 (Monterey), a regression introduced in 2.1.255
Fixed remote and scheduled sessions failing with a "non-empty content" error after a re-sent permission approval

Claude Code v2.1.257

Tue, 01 Sep 2026 00:00:00 GMT

Added Claude Fable 5.1 (claude-fable-5-1), now the default Fable model: 1M context, $10/$50 per Mtok
Added a Containment Escape rule to auto mode against cloud metadata-credential fetches, egress evasion, and cross-tenant reach
Added a one-time auto-mode prompt before the first file read outside the working directories (permissions.blockReadsOutsideWorkingDirectories)
Changed defaultMode: "bypassPermissions" in project/local settings to be ignored, like "auto"

[Guide Release] Claude Code Ultimate Guide v3.43.0

Tue, 01 Sep 2026 00:00:00 GMT

A shorter repository entry point now shares one Start, Build, Scale, Resources, and Updates navigation contract with the public sitemap. This release also includes the published MCP 1.3.1 package and Registry entry, loop and graph engineering, subscription strategy, translation governance, and the current security and reliability research accumulated since v3.42.0.

Claude Code v2.1.252

Mon, 31 Aug 2026 00:00:00 GMT

Fixed Bash commands failing with a task-output-swap error on some Macs
Fixed "always allow" not saving in a project with no .claude/settings.local.json yet
Fixed Remote Control sessions hosted by Claude Desktop or VS Code stalling for minutes after a degraded claude.ai connection

[New Section] AI Coding Cost Controls at Scale

Mon, 31 Aug 2026 00:00:00 GMT

AI Unit Economics now separates request-, task-, and delegation-level routing, progressive spend gates for interactive developers, terminal budgets for unattended workloads, and cache effects. Databricks internal routing and token-reduction results remain attributed operating evidence rather than portable savings benchmarks.

[New Page] Translations and Language Status

Mon, 31 Aug 2026 00:00:00 GMT

New public status page for the canonical English guide, the maintained French edition, and independent Chinese and Ukrainian community translations. It records versions, synchronization dates, source provenance, coverage, and the current translation gaps without presenting community projects as official editions.

[New Page] Loop & Graph Engineering

Sun, 30 Aug 2026 00:00:00 GMT

New canonical page for designing repeated feedback loops and executable workflow graphs without smuggling judgment into opaque automation. Covers nodes, edges, joins, checkpoints, shared state, stopping rules, and where human judgment remains explicit and accountable.

[New Page] Subscription Strategy at Team Scale

Sun, 30 Aug 2026 00:00:00 GMT

New ops page for choosing between Claude Team seats, usage-priced Enterprise, a governed API gateway, multi-vendor setups, and self-hosted inference. Documents the verified Claude Team 2-150 seat cap, current billing and contract boundaries, observed cost distributions, timestamped practitioner reports, and scenario-bound self-hosting economics.

Claude Code v2.1.251

Fri, 28 Aug 2026 00:00:00 GMT

Added PreModelSwitch and PostModelSwitch hook events (block, confirm, or annotate a model switch); SessionStart resume hooks now receive session staleness and the estimated re-cache cost
Added live streaming of a foreground subagent's tool calls and results to Remote Control clients
Added a spend limit bar to /usage and a rate_limits.spend_limit status line field for gateway users
Fixed three security gaps: a symlink swap bypassing file-tool permission checks, plugin marketplace command paths escaping the plugin directory, and sandboxed Bash output-file redirection
Changed CLAUDE_CODE_SUBAGENT_MODEL to set the default subagent model rather than override everything; an agent definition's model: and an explicit per-spawn model now take precedence over it

[New Section] Agent Harness Map: Runtimes, Orchestrators and Meta-Harnesses

Fri, 28 Aug 2026 00:00:00 GMT

The evidence-backed map now covers 192 projects: a pinned 160-project source directory plus 32 directly researched guide supplements. It separates 42 strict runtimes from 15 adjacent control planes, adds an evidence-pinned Liza profile, and keeps a separate research layer for harness optimizers and meta-harnesses.

[New Recap Card] C14 Recap Card: Agent Harness Map

Fri, 28 Aug 2026 00:00:00 GMT

A one-page A4 decision aid for separating the model, runtime, repository contract, and orchestrator, then testing a shortlist on evidence and accepted-task cost.

Claude Code v2.1.250

Thu, 27 Aug 2026 00:00:00 GMT

Bug fixes and reliability improvements

Claude Code v2.1.248

Thu, 27 Aug 2026 00:00:00 GMT

Added --restricted (or CLAUDE_CODE_RESTRICTED=1): removes the built-in tools that run commands or code and WebFetch (unless named in --tools), keeps file tools inside the working directory, refuses bypassPermissions, and ignores user, project, and local settings files
Added cross-session messaging (SendMessage/ListAgents) between sessions on the same machine on Bedrock, Vertex, and Foundry, and when telemetry is disabled
Fixed a prompt-cache miss (and lost extended-thinking context) roughly once an hour in long sessions, caused by tool definitions being re-rendered after an OAuth token refresh
Added experimental.cacheTtl ("5m" or "1h") to agent frontmatter for a per-agent prompt cache TTL

[New Page] Monitor, Channels and Safe Event Delegation

Thu, 27 Aug 2026 00:00:00 GMT

New workflow for choosing between Monitor command and WebSocket sources, plugin monitors, MCP Channels, and Routines. The GitHub-to-Codex path verifies and deduplicates events, starts read-only, and requires an explicit gate plus an isolated worktree before any write-capable task.

[New Page] AI Executive Agents: Virtual C-Suites and Board Simulators

Thu, 27 Aug 2026 00:00:00 GMT

New ecosystem page mapping AI agents that simulate named executive roles (CFO, CMO, CHRO, General Counsel, board of directors). Deep dive on OpenExecutive's 8-agent architecture, five open-source alternatives with GitHub-API-verified star counts (most in single digits, none close to Gstack's traction), and a routing table by role. Closes on the augmentation-vs-replacement question and the theoretical Synthetic Director legal framework.

Claude Code v2.1.247

Wed, 26 Aug 2026 00:00:00 GMT

Added the SendFeedback tool: Claude can draft a feedback report for you to review and send from /feedback
Changed Sonnet 5's default auto-compact window to its full 1M context, auto-compacting at about 967K tokens instead of about 934K
Added /claude-api cost-optimize to profile an existing project's Claude API spend and work through cost levers
Fixed sub-agents dying on a first-call model 404: they now use the session's fallback model chain

[Guide Release] Claude Code Ultimate Guide v3.42.0

Wed, 26 Aug 2026 00:00:00 GMT

Evaluated Anthropic's AI-native SDLC playbook blog post and closed the two real gaps against the guide's existing spec-first documentation: an intent.md upstream problem statement, gated by a PM before the technical spec starts, and a closed Maintain-to-Plan loop where a production monitoring threshold drafts a new intent.md automatically. Documented in spec-first.md's 3-document chain (intent.md, spec.md, plan.md) and the Spec-First Development Pipeline diagram.

Claude Code v2.1.246

Tue, 25 Aug 2026 00:00:00 GMT

Added a startup warning for Bash allow rules with a wildcard before the subcommand (e.g. Bash(git * main)), since they also match options inserted before the subcommand
Added an Auto mode tab to /permissions for viewing and editing auto mode classifier rules
Fixed the Write tool reporting "Out of memory" or freezing for a long time after overwriting a very large existing file, even though the file had been written
Improved non-interactive sessions (-p, SDK, cloud sessions) to automatically continue a response cut off mid-stream by a server error, connection loss, or stall

[Guide Release] Claude Code Ultimate Guide v3.41.3

Thu, 20 Aug 2026 00:00:00 GMT

Consolidates several weeks of security, agent-harness, and sandbox research: the threat database jumps from v2.20.0 to v2.27.0 across the Shai-Hulud keyv npm worm, the Ghostjacking log-poisoning technique, and the skills.sh Paperclip typosquat campaign. New agent-harness-landscape.md page compares 25 CLI, IDE, and cloud harnesses. Sandbox troubleshooting rewritten around what actually breaks in daily use, not what looks dangerous on paper. The Claude Code release tracker catches up to v2.1.237, and the /ultraplan section is removed: the command was deleted from Claude Code five months ago and the guide never noticed.

[New Section] Why reviewing SKILL.md stopped being enough

Mon, 17 Aug 2026 00:00:00 GMT

Threat DB v2.27.0 adds seven primary-source-verified CVEs. CVE-2026-54316 affects Claude Code from 0.2.54 before 2.1.163: huggingface.co was pre-approved as a bare hostname for WebFetch, creating an out-of-band data channel through attacker-controlled paths. CVE-2026-12537 affects Gemini CLI and run-gemini-cli in headless CI before their 0.39.1 and 0.1.22 fixes. The new guide subsection covers the skills.sh Paperclip campaign: reviewing only SKILL.md missed instructions in a secondary file. An immutable commit or digest pin protects reviewed content from silent mutation, but every pin change still needs a fresh review.

[New Section] Sandbox Isolation: just-bash, the pure-simulation counter-example

Tue, 04 Aug 2026 00:00:00 GMT

New entry in the Cloud Sandboxes Landscape section: just-bash (Vercel), a bash interpreter written directly in TypeScript with no VM at all, the opposite trade from agentOS a few lines above it. Same in-process, no-cloud-account category, weaker containment, zero moving parts in exchange. Named directly: no LICENSE file at the repo root despite the README claiming Apache-2.0.

[New Page] New workflow: Agentic Software Factories

Wed, 15 Jul 2026 00:00:00 GMT

The topic lived across seven files with no entry point, so this page is the map. It carries a six-level spectrum from a single Claude Code session up to a closed software factory, with the cost and the upgrade signal at each step, plus the decision tree that did not exist: when a paid factory actually beats native Agent Teams and /batch, which already cover most of the same need. Also a five-question governance checklist and a case study on what unbounded agentic velocity produces.

[New Page] Role pages: PM, Tech Lead, CTO, CIO/CEO

Mon, 06 Jul 2026 00:00:00 GMT

Four audience pages are now readable online: For Product Managers, For Tech Leads, For CTOs, and For CIOs & CEOs. Each one answers the same question for a different role: what Claude Code changes for you and where to start.

[Guide Release] Claude Code v2.1.186

Tue, 23 Jun 2026 00:00:00 GMT

New claude mcp login/logout command authenticates MCP servers directly from the CLI without the interactive /mcp menu, with --no-browser support for SSH. ! bash commands now auto-trigger Claude to respond to output (opt-out via respondToBashCommands: false). Background subagents surface permission prompts in the main session instead of auto-denying. Fixed Agent(type) deny rules not enforced for named subagent spawns. 25+ bug fixes.

[New Page] Team Knowledge Infrastructure - shared KB for Claude Code + Cowork

Tue, 23 Jun 2026 00:00:00 GMT

New guide page: 3-tier framework for setting up a company knowledge base accessible from both Claude Code and Cowork. Covers the versioned Markdown vault pattern (direct file reads, ~/Shared/CLAUDE.md bridge), MCP connectors for live systems (Atlassian Jira+Confluence, Notion, GitBook), RAG at scale with the ~100-1000 doc threshold, Onyx (self-hosted), LlamaCloud and Ragie (managed via MCP), and the plugin pattern for distributing team workflows. Includes governance caveat on Cowork audit log and DLP gaps.

[New Page] Practitioner Insights - field reports across six source corpora

Sun, 21 Jun 2026 00:00:00 GMT

Expanded guide page: 69 paraphrased insights from IFTTD, Devoxx, Dev With AI Meetup, ByteByteGo, Stanford Online, and Pavan Belagatti's 62-video 2026 corpus. Covers context engineering, agentic patterns, LLM evaluation, agent security, DevX, and adoption with explicit evidence boundaries.

[New Page] New workflow: Smart-Suggest BM25 Skill Routing

Fri, 19 Jun 2026 00:00:00 GMT

New workflow guide documenting a self-calibrating BM25 lexical scoring hook for UserPromptSubmit. Scores every prompt against a curated skill corpus (positive/negative examples per skill), auto-calibrates per-skill thresholds via F-beta leave-one-out cross-validation, and injects ranked suggestions as additionalContext. Complements regex-based routing with broad natural-language coverage. Includes a runnable Node.js example with zero npm dependencies.

[Guide Release] Claude Code v2.1.183

Fri, 19 Jun 2026 00:00:00 GMT

Auto mode now blocks destructive git commands (reset --hard, checkout -- ., clean -fd, stash drop) and git commit --amend (for commits not made by the agent) and terraform/pulumi/cdk destroy unless you explicitly asked. Model deprecation warnings added on stderr in print mode and for agent frontmatter models. New attribution.sessionUrl setting omits the session link from commits/PRs. /config --help lists all shorthand keys. 10+ bug fixes including thinking 400 errors, WebSearch in subagents, and tmux pane launch failures.

[Guide Release] Claude Code v2.1.181

Thu, 18 Jun 2026 00:00:00 GMT

New /config key=value syntax lets you set any setting inline from the prompt (e.g. /config thinking=false), working in interactive, -p, and Remote Control modes. Long paragraphs now stream line-by-line instead of waiting for the first line break. Subagent panel improvements: idle agents auto-hide after 30s, list caps at 5 rows with scroll hints. 30+ bug fixes including prompt caching on custom ANTHROPIC_BASE_URL, Write/Edit truncated files on network drives, and a 120ms startup regression.

[Guide Release] Claude Code v2.1.178

Wed, 17 Jun 2026 00:00:00 GMT

New Tool(param:value) syntax for permission rules lets you match tool input parameters with wildcards, e.g. Agent(model:opus) to block Opus subagents. Nested .claude/skills directories now load automatically with name-clash disambiguation. Auto mode subagent spawns are evaluated by the classifier before launch. 10+ bug fixes including subagent transcript visibility, compaction fallback model, and MCP disallowedTools for server-level specs.

[Guide Release] Claude Code v2.1.176

Mon, 15 Jun 2026 00:00:00 GMT

Session titles are now generated in the language of your conversation, with a new `language` setting to pin a specific one. New `footerLinksRegexes` setting for custom footer badges. New `enforceAvailableModels` managed setting (v2.1.175) locks the Default model to the allowlist. 15+ bug fixes across Remote Control, tmux clipboard, hook path conditions, and Bedrock credential caching.

[Guide Release] Claude Code v2.1.172

Thu, 11 Jun 2026 00:00:00 GMT

Sub-agents can now spawn their own sub-agents up to 5 levels deep. Bedrock auto-detects the AWS region from ~/.aws config files. 20+ bug fixes including the 1M context stuck issue and model picker availableModels filtering.

[Guide Release] Claude Code v2.1.170

Tue, 09 Jun 2026 00:00:00 GMT

Claude Fable 5, a Mythos-class model exceeding any previously generally available Anthropic model, is now accessible in Claude Code v2.1.170.