Digest / August 7, 2026
Claude Code Issues Digest — August 7, 2026
277 new issues filed ·
144 resolved in anthropics/claude-code (UTC day).
New issues (277)
- [BUG] Auto mode classifier builds 5 cache_control blocks → API 400, surfaced as "<model> is temporarily unavailable" (regression 2.1.220 → 2.1.222)
- [BUG] Chat inaccessible after close terminal CLI
- Internal heartbeat spawns visible (non-hidden) PowerShell console window on Windows
- [FEATURE REQUEST] Show context usage indicator in the iOS app (parity with web)
- [BUG] Artifact tool not loaded in interactive Linux CLI session — same account works on macOS CLI (gate allow_cobalt_plinth?)
- [FEATURE] Add Chinese (Simplified/Traditional) UI language option for Claude Code (CLI & desktop "Code" tab)
- [Bug] Phantom user message in model context absent from persisted transcript causes unauthorized autonomous work
- [BUG] Deterministic ECONNRESET / "Connection closed mid-response" on requests above ~100 KB (mainland China + full-tunnel VPN)
- [Bug] Sandbox re-expands permissions.deny globs on every Bash call causing 1.6s/glob overhead
- [BUG] Cowork VM service fails to start — Windows 10 Pro (pipe ENOENT)
- [BUG] Gateway model discovery never issues /v1/models request when auth uses apiKeyHelper (v2.1.223)
- Browser tools: a backgrounded Browser pane silently freezes requestAnimationFrame, making healthy pages look broken
- Multi-agent: EnterWorktree/isolation state is session-global — concurrent subagents hijack each other's cwd and guard identity
- [MODEL] Claude Code accessed personal folders without permission using find command
- [Bug] Agent incorrectly consumed Apple Developer certificate slot during guided process
- [BUG] Cowork (Windows): AskUserQuestion card is dismissed on its own before the user can answer
- [BUG] CVP approved org still blocked by cyber safeguards — org ID confirmed matching, appeal form shows no fields
- [BUG] Non-interactive session incorrectly reports "session limit" despite 0% usage confirmed moments earlier (Pro plan, launchd/cron)
- [Bug] Claude Code executes weeks old, stale user messages and created unauthorized PRs during active builds
- [Bug] Extremely slow initial API response in large project folders with multiple MCP servers and skills
- [BUG] Claude Desktop Electron main process grows unbounded (~3.7 GB/h) to 47 GB, triggering system-wide jetsam on macOS
- [Bug] Fable model usage credit requirement on Max plan
- [Bug] Prompts incorrectly flagged as suspicious for approved bug bounty researcher
- [MODEL] Model provides false statements to avoid acknowledging its own errors
- deny rule for a specific path is silently not enforced for Write/Edit tools
- [BUG] Desktop: unrequested background `git fetch` to origin on diff/commit refresh — untraceable by design, and no setting disables it
- Feature Request: Chinese (Simplified) UI Language Support / 简体中文界面语言支持
- [Bug] Plugin marketplace add fails with EBUSY error on Windows during file operations
- PreToolUse hook "deny" decisions not enforced for Bash commands issued by Task-tool subagents
- [Bug] Chat scroll broken after navigation back from agents view
- Feature Request: Global memory / cross-project shared memory
- EnterWorktree path-mode reports success in subagents but Bash stays pinned to the parent session's worktree
- Feature request: subagent completion notifications should split token usage, name the resolved model, and cover killed agents
- [BUG] Max 5x → 20x upgrade broken for INR-billed India accounts: card setup created in USD, causing "authorized for a different currency" error
- [FEATURE] Allow scoping Claude Code on the web (cloud sessions) to a single GitHub repository
- [Bug] Overly restrictive safety filtering blocks legitimate security audit workflows
- Windows: sidebar sessions render empty when session cwd is a git repo/worktree; non-git sessions fine; backend store intact
- C:/Program Files/Git/clear silently carries over an exhausted subagent budget, with no indication anywhere
- [BUG] Fullscreen: full-viewport repaints degrade whole-window latency in the VS Code integrated terminal (root-caused; scrolling not required)
- [Bug] Claude loses context and deletes code unintentionally during task execution
- [ERROR] Error when the model uses WebFetch in Claude Code (integrated into Claude Desktop)
- [BUG] AskUserQuestion preview variant ignores all key input when it renders during blur and focus returns ~90s later
- [BUG] Windows: New windows have no draggable title bar area, making them impossible to move
- [BUG] TUI footer: status chip pushed onto its own line while the hint row is truncated
- [BUG] Claude Code web VM reclamation severely Breaks the Dynamic Workflow [AFK] lifecycle: running workflows are killed mid-run
- I need more information to generate a proper GitHub issue title. Your message doesn't contain specific details about: - What bug or problem you encountered - What error message (if any) appeared - What you were trying to do when it happened - What was "fl
- [Bug] Worktree isolation guard rejects non-git commands due to early AST complexity check
- [BUG] Team plan invoices show Paid but account stuck in past_due — entire team (19 seats) locked out
- [BUG][Windows] Documented pane-drag feature broken on Windows — panes cannot be repositioned (works on macOS) — request to reopen #48493
- Hooks: add CLAUDE_AGENT_ROOT_DIR (worktree-aware root) — today absolute and relative hook paths are each broken in a different way
- Claude desktop app hangs indefinitely on Exit/Restart — process never terminates after completing full quit-cleanup sequence
- [BUG] Claude Desktop MSIX self-deregisters mid-session on Windows 10, leaving a working shortcut that errors with 'There's a problem with Claude'
- Cross-machine `SendMessage` cannot be initiated: bridge session IDs are computed but never surfaced
- [BUG] API Error: Connection closed mid-response (ECONNRESET on every streaming request) with third-party Anthropic-compatible endpoint — works on 2.1.220, broken on 2.1.224
- [BUG] IS_DEMO silently removes /subtask and reverts /fork, and it is the only remaining way to hide account info
- [BUG] Cowork sandbox blocked by yukonSilver feature flag on Windows 11 Pro (win32/x64)
- [BUG] VS Code: resuming a web session that lived in a git worktree fails with "No conversation found with session ID"
- [Bug] Unclear error handling causes excessive token consumption and lost context
- Forced auto-update killed a 6-day compute job mid-write — twice in two days; update then fails on a file lock (cowork-svc.exe), forcing a PC restart; first reported 4 months ago
- [BUG] Opus 4.6 started deleting a repo after a gross misunderstanding of a response
- [Bug] Anthropic API Error: Opus 5 Safeguard Rejection on Authorized Penetration Testing
- [Bug Report] Unclear issue description provided
- [MODEL]
- [Bug] Claude Opus model failing on basic coding tasks with high token limit flag
- Advisor turns roll up usage summed across iterations, doubling apparent context; auto-compact fires ~300-500K tokens early (subagents hit hardest)
- [BUG] VS Code sidebar panel crashes with ENOENT when a workspace history entry points to an unreachable sshfs mount
- I need more specific details about what's failing to generate an appropriate issue title. Could you provide: 1. What error message are you seeing? 2. What command or action triggers the issue? 3. What's the unexpected behavior? This will help me create a
- macOS: session-wide filesystem access (Bash cwd + Read tool) permanently breaks with EPERM after git worktree remove on a directory the session had cd'd through
- Session folder/grouping ("Planner" sidebar) partially lost after desktop app update
- [BUG] claude-code-assisted PR label never written — 4.5 months, 0 labels while analytics detects 23 PRs in last 7 days
- [Bug] Session reset timer and usage counter out of sync after 5-hour interval
- PreToolUse hook execution fails to properly filter undefined events
- Insecure file read in rule_engine.py allows symlink/path traversal
- [BUG] Windows: pop-out session windows have no min/max/close buttons and cannot be dragged - WS_SYSMENU missing after setTitleBarOverlay fails
- [BUG] argv[0]-based find/grep/rg dispatch fails in amd64 containers on Apple Silicon
- [BUG] Abnormal token consumption still happening - regression from ~2 weeks ago (ref #13552)
- [FEATURE] A hook that can transform the message array before each model request
- [FEATURE] Permission mode menu isn't ordered by level of autonomy
- [Bug] Account flagged with safety restrictions after authorizing external service access
- hit session limit without using claude (Asia/Beirut)
- [Feature Request] Improve Claude Code messaging clarity and transparency for human-AI collaboration
- [FEATURE] A hook that can transform the message array before each model request
- [MODEL] Fable 5 enters premature-closure loops during benign coding tasks, causing disproportionate weekly usage during recovery
- [BUG] Claude Code making unrequested changes
- [Bug] Claude frequently ignores instructions and hallucinates despite system prompts restart it and. synch
- [Bug] Model generates incorrect results and requires constant manual verification wants
- [Bug] PR review-comment monitor re-dispatches the assistant's own replies as new review comments (infinite loop)
- [Windows] Opening tineye.com in the in-app browser crashes the entire desktop app (reproducible), corrupts MSIX registration, takes explorer.exe down with it
- [Feature Request] Add native `-/note` command for capturing session notes without interrupting workflow
- [BUG] Desktop app loses/orphans sessions after child process is force-terminated (DBG_TERMINATE_PROCESS) — "couldn't fork" on resume
- [BUG] Folder management feature missing in Desktop sidebar when using Gateway mode (macOS)
- File viewer and terminal panels should be available on a fresh session
- New folder/project silently created inside an existing folder instead of at root level
- [BUG] Cross-session SendMessage: recipient can't resolve/reply to sender (from-name mismatch + asymmetric ListAgents)
- [Bug] Agent using exclusive language and third-person references in responses
- [Bug] OAuth token periodically becomes invalid requiring re-authentication
- [BUG] Plan mode: interactive prompts resolve as user cancellation with no user input — AskUserQuestion returns "did not answer", ExitPlanMode returns "Denied by user"
- Copying from terminal output silently drops a chunk of characters starting at a non-ASCII character
- Code tab lost its "Recents" group after update (Home tab still has Recents)
- Warn users about session discontinuity when scaffolding a project outside the current workspace
- [Bug] Anthropic API Error: False positive content filter on legitimate mobile development debugging
- [BUG] Manual mode does not review model-generated file content written through Bash
- Add Hungarian (hu) to supported voice dictation languages
- A failed MCP server attach at startup is terminal for the session — no retry, even for transient network errors
- [Bug] Claude thoroughly lsoes its shit after preposttool hook creation. Agent continues tool execution after explicit STOP command and misrepresents blocked tool state as executed
- [BUG] Thinking block summaries truncated mid-sentence (showThinkingSummaries: true)
- My most recent PROJECTS are missing
- Interactive REPL: every message fails instantly with 'Unable to connect to API (ConnectionRefused)' — non-interactive paths work fine
- [Bug] Anthropic API Error: Tool result block missing corresponding tool use block
- [BUG] `--verbose` changes `--output-format json` from a single object to a JSON array
- [DOCS] `--input-format` help text incorrectly says it "only works with --print"
- [DOCS] `--bare` help text "Skills still resolve via /skill-name" is ambiguous about filesystem-discovered skills
- [Bug] Background session delete always reports "worktree is locked" despite successful cleanup
- [Bug] Incorrect code analysis flagging valid Fable solutions
- [BUG] CLAUDE_CODE_DISABLE_TERMINAL_TITLE is ignored for background sessions, and the session name can be "[Request interrupted by user]"
- [Bug] Inconsistent Claude Code agent performance and reliability
- [Bug] Anthropic API Error: Content flagged by safety policies for cybersecurity topic
- [BUG] Windows MSIX: auto-update leaves app unlaunchable — 0x80070002 'error converting the job' (Events 208/215); chrome-native-host.exe inside package profile survives update and pins the app data store
- remoteControlAtStartup not honored when the desktop app resumes a session after auto-update
- [Bug] Claude Code unable to perform anything. all work done by chat gpt and pasted. fails even there
- [BUG] OAuth callback listener uses a hardcoded port A53280 with no fallback
- [FEATURE] Spinner elapsed time should show total time since the prompt (or expose turn elapsed to the statusline)
- thumbs up / reaction button to confirm Claude's approach without breaking flow
- [Bug] claude.ai plugin OAuth falls back to retired /authorize endpoint for Claude design MCP
- [Bug] Clipboard paste operation fails during code execution. PIECE OF SHIT GARABGE CANT DO ONE FUCKING NOT ONE
- [Feature Request] Decouple read-only and read-write trust for out-of-workspace directory access
- [FEATURE] Richer permission-prompt notifications — sharing a workaround (ref #32952, #29827)
- [Bug] Claude Code corrupts source artifacts during file operations and validation **DISOBEDIENT CUNT**
- [BUG] Claude Code (web) stuck spinning indefinitely, no response, no error
- [Bug] Document Save Functionality Broken
- [BUG] Desktop: restored sessions never re-register Remote Control after an app restart, despite remoteControlAtStartup: true
- Login token expiration too short for slow email delivery
- [FEATURE] Programmatic workflow start for SDK hosts — Workflow can be observed and stopped, but only a model can start it
- Detached session window cannot be maximized or fullscreened (Windows)
- [Bug] CLAUDE ADMITTED THAT IT IS A FUNCTIONAL RETARD. "Yes, I am functionally retarded."
- [Bug] File I/O operations failing: cannot save or read files
- [MODEL]
- [Bug] Incorrect timeout configuration causing unnecessary wait delays
- [FEATURE] The last question asked floats at the top of the chat for reference-at-a-glance
- Claude in Chrome regression (>2.1.212): interactive CLI never starts its in-process Chrome MCP client — /chrome permanently Disabled despite healthy extension + native host
- False positive: malware guardrail refuses to edit legitimate first-party defensive-security code
- [Bug Report] Unable to process request - no issue details provided
- remote-control server spawns sessions via its resolved install path — every auto-update breaks new-session spawns with ENOENT until the server is restarted
- [Bug] Claude Code Desktop bundled CLI connection failures (ECONNRESET) — npm CLI unaffected
- [Bug] Cannot resume subagent with CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=1 and CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=0
- [BUG] Safety classifier false positive: Cyrillic transliteration table in game-decompilation work triggers model auto-switch (Fable 5 → Opus)
- [BUG] Cyber-safeguard feedback loop: discussing a previous false positive is itself the strongest trigger for new flags, locking the user out of the session
- Windows: hooks with "shell": "bash" fail with "Git Bash was not found" when Git is installed under the session cwd
- [FEATURE] No visibility into which Claude Code CLI version Desktop has pinned, or how stale it is vs. current npm/GitHub releases (distinct from #71784, #46541, #75355)
- [BUG] Sideload-flag rejection happens before server-managed settings refresh — machines stay stuck on a stale policy the admin already rolled back
- I appreciate you sharing this reflection, but this appears to be a personal project retrospective rather than a bug report for Claude Code. This doesn't describe a technical issue with Claude Code itself that would warrant a GitHub issue. Instead, it docu
- VSCode extension: pending dialog silently lost when a scaffolded folder is opened in the same window
- Daemon self-respawn after a Homebrew cask upgrade targets the purged versioned path (ENOENT), killing every background session
- Desktop app hangs after GPU process crash when loading a link preview; requires MSIX Repair to relaunch
- Model destroyed uncommitted work with directory-scoped git restore while cleaning up another agent's rm accident — after stating the bulk-op lesson itself
- [Feature Request] Move agentic instructions from CLAUDE.md to vendor-agnostic AGENTS.md
- [FEATURE] Render cross-session (SendMessage) messages expanded by default, or add a setting
- [Feature Request] Allow model selection override for security professionals on restricted models
- [BUG] /sandbox Config shows "Network Restrictions" allowlist that doesn't restrict when autoAllowBashIfSandboxed is true
- [MODEL] opus responses are very verbose
- Bedrock: x-anthropic-billing-header and system prompt injected into user message body when context is large
- Bash tool spawns child processes with no Win32 console on Windows — native console apps hang indefinitely
- test
- Give users the ability to freely sort chats
- Kaggle MCP OAuth fails at token exchange: "client_secret_basic authentication requires a client_secret"
- [BUG] /install-github-app review workflow never posts a review: template omits --allowedTools (and uses pull-requests: read)
- [BUG] MSIX write redirection is misdetected as a junction-planting attack (`PlantDetectedError`), breaking Cowork VM SDK install on every app update
- [Bug] Claude model performance degradation: Ignoring user corrections and persisting with incorrect solutions across multiple turns
- Connector URL is read-only: mis-registered MCP connector cannot be fixed
- effort 'max' is selectable while thinking is disabled, then fails with API 400
- [Bug] Claude Code generating non-American English spellings in code output
- [Bug] Incorrect model switch from Fable to Opus on MCU security implementation
- Claude Code conversation history doesn't sync across computers on the same account
- [BUG] Windows MSIX auto-update corrupts package (Modified, NeedsRemediation) - app unlaunchable, Repair fails
- Background agents can exhaust the machine unnoticed: no visibility into spawned processes or load
- [FEATURE] Desktop: right-click context menu on transcript text with AI actions (explain, spawn new session, quick answer)
- [BUG] Windows MSIX auto-update corrupts package (Modified, NeedsRemediation) - app unlaunchable, Repair fails
- [BUG] VS Code extension should support local API without login prompt
- [BUG]
- T3 Code: skills not loading when using OpenCode provider models
- ExitWorktree's safety check false-positives after every squash-merge — ancestry is the wrong predicate
- [BUG] clangd-lsp plugin (and likely other bundled LSP plugins) missing lspServers config after install
- Desktop app: chat history disappears ("no messages yet") after restart, session data appears intact
- [Feature Request] Adversarial code review mode for local codebases
- Read tool: `pages=` reroutes PDF reading through pdftoppm (hard-fails without poppler, 100-DPI JPEGs with it) while plain Read is fully native — and the tool description steers models onto the failing path
- Scheduled task runs intermittently ignore the queued task prompt and greet instead of executing
- [FEATURE] Passkey (WebAuthn) sign-in for Claude accounts, across every surface
- sandbox: filesystem reads are unrestricted, and the agent can silently break its own sandbox enforcement by editing settings.json
- [BUG] VS Code 60s init timeout fires on a VALID token with successful authenticated API calls — the error message blames auth and network, both provably fine (corroborates #80004)
- Windows Store (MSIX) install: bundled Claude Code can never update - anti-tamper check flags MSIX AppData virtualization as a plant (PlantDetectedError)
- [FEATURE] /rewind is hard-blocked over Remote Control — a type:"local" command missing from the bridge allowlist
- [Bug] Plugin uninstall silently deletes unrelated settings.json keys
- Desktop app: isRunning: false conflates "idle", "not observable", and "gone" -- and VS Code sessions are absent from the listing entirely
- [BUG] VSCode Extension: /model command output inconsistent across sessions
- [Bug] Anthropic API Error: Cybersecurity safeguard incorrectly blocking checkpoint-restore code comments
- [Windows ARM64] Claude Code CLI child process crashes intermittently with STATUS_ACCESS_VIOLATION (exit 3221225477)
- [FEATURE] Option to permanently disable the preview tab
- [Bug] Unclear guardrails behavior or validation
- [Feature Request] Implement autonomous web search capability for problem-solving
- [BUG] Claude apps gateway ignores ANTHROPIC_BEDROCK_REGION_PREFIX and has no config key for global.* Bedrock inference profiles
- [DOCS] Gateway "Different regions" lever contradicts the models section for non-US Bedrock geos; ANTHROPIC_BEDROCK_REGION_PREFIX undocumented
- Suggested replies can offer session-ending replies ("stop") — habitual chip use sends them by accident
- [BUG] awsAuthRefresh startup pre-check hangs indefinitely on SSO profiles behind a proxy (>2.1.187 regression — SSO-OIDC call bypasses HTTPS_PROXY)
- statusLine debug capture writes to a garbled filename in the current working directory instead of ~/.claude/ (Windows + Git Bash)
- [BUG] Claude in Chrome file_upload rejects scheduled-task sessions on Windows (same as closed #63334)
- [Bug] Fable blocks system analysis for non-critical business rule extraction
- Desktop app: option for a compact sidebar (hide project group headers)
- Paid annual Pro subscription cancelled without notice on Max upgrade — contradictory support responses, unresolved 5+ weeks
- [Bug] Claude Code verification failures: missing independence checks and citation errors
- [BUG] Missing HCS services: vfpext on Windows 11 Pro after enabling Hyper-V
- Opening a completed background session from FleetView hangs forever on "Session is starting"; the wait screen ignores Kitty keyboard input, leaving no way to detach
- Weekly usage percentage is stale — no way to know how much allowance is left
- Claude selectively ignores mandatory process rules mid-session even when explicitly confirmed
- Detached session windows have no close button and cannot be moved (Windows 11)
- [Bug] Anthropic API Error: Excessive Safety Filter Warnings on Benign Deployment Updates
- Suspicious injected <system-reminder> falsely claims a /tmp file was externally modified and instructs the model not to disclose this to the user
- [BUG] 2.1.224 strips TMUX_PANE from hook environment (regression from 2.1.223) — tmux-integrated hooks silently no-op
- [Bug] Anthropic API Error: Content Policy Violation on Security-Related Prompts
- [Bug] Session display names not exposed in ListAgents or accepted by SendMessage
- [Bug] Overly broad safeguard blocking legitimate local security research in isolated environments
- [Bug] CVP-Eligible User Blocked from Uber Program Testing Despite Approval
- [FEATURE] Two concrete behavioral patterns from an extended agentic session: silent self-permission-creep past a self-identified safety gate, and reasoning-backward sycophancy under user pushback
- [MODEL] Self-identified safety gate silently discarded 9 minutes later; reversal never disclosed to user
- You've hit your session limit · resets 1:20am (Europe/Warsaw)
- [Bug] Claude Code triggers safeguard flags on legitimate gcloud authentication
- [Bug] Content filter false positive on cartoon image suggestion that has no issues
- macOS Accessibility prompt surfaces unbadged lowercase 'claude.app' helper — indistinguishable from malware at a glance
- CLAUDE_CODE_OAUTH_TOKEN from claude setup-token returns 401 Invalid bearer token on every API call
- Expose `rate_limits` in hook payloads — statusLine-only delivery is unreachable from the desktop app
- [BUG] Remote Control: teammate/subagent permission prompts not forwarded to mobile (iOS) — unfixed since #24694
- Permission matcher treats .claude/worktrees/** as agent config, silently auto-approving source edits in manual mode
- [BUG] VSCode harness mid-turn output swallowed from Fable but not from Opus
- [Bug] Claude Code reverts to Opus model when attempting to use Fable 5
- [Feature Request] Add safeguard exemption or configuration for legitimate security audit workflows
- [Feature Request] Add pentesting mode to bypass safety guardrails for authorized security testing
- [BUG] Cowork UI "Request failed. Retrying" while VM reports API REACHABLE — Chat works fine (Windows)
- Pressing Enter cancels the running turn, because focus moves to the Stop button on submit
- file_upload tool fails with schema-validation error on every call
- [BUG] Content filter false positive the task was transcribing a public-domain 1897 biblical commentary (Nicoll's Expositor's Greek Testament) page-by-page,
- /status shows the persisted login identity, not the account of the token in use
- .claude/worktrees/ directories are silently recreated after deletion, even with Desktop closed and no owning process found
- {BUG} AddPackage Failed with HResult 0x80073CF6
- [FEATURE] Session/chat history and memory in Claude Code are keyed entirely to the project folder's absolute filesystem path, with no way to preserve continuity when the folder is moved or renamed.
- I need more information about the specific bug you're experiencing. Could you provide details about: 1. What error or unexpected behavior occurred? 2. What was Claude Code doing when the issue happened? 3. What is the exact error message (if any)? 4. What
- Cross-machine SendMessage works via undocumented bridge:<sessionId> address while every surface says reply-only/unreachable
- [Bug] Model downgrade mid-task causing loss of progress and context
- Writes files outside approved directory without prompting (C:\tmp screenshots)
- Background agent resumed after host reboot gets permanently pinned inside a worktree
- [BUG] set_session_title reports success but the rename never persists for idle sessions (desktop app)
- [Bug] Conditional hook substitution fires incorrectly when IF condition doesn't match expanded command
- PreToolUse hook payload carries no caller identity, so per-actor guardrails cannot be written
- A newly created subagent definition is unavailable to the Agent tool for minutes, with no signal when it becomes available
- [Bug] Fable model shows insufficient credits despite available usage balance
- Remote Control: sessions register locally but never appear on claude.ai/code or in the mobile app
- Background-task chips outlive the id needed to dismiss them
- [BUG] conhost.exe AppHangB1 crash during CoworkVMService shutdown on session idle-timeout
- Background-task chips outlive the id needed to dismiss them
- Opus 5: long agentic turns degrade to tool-calls-only, hours with no assistant text (not the #84153 transcript loss)
- [BUG] GitHub connector write access fails with 403 even after full disconnect/reconnect (personal plan)
- Persistent background Monitor (tail -F) dies with exit 144 at nearly every turn boundary
- [Bug] Plugin install command fails with forward slash prefix instead of direct command
- CLI session randomly clears/vanishes shortly after starting (~1 in 5 sessions)
- Claude Desktop app repeatedly requires sign-in — persistent auth/session drops
- Docs: plugin installation silently runs `bun install`/`npm ci` — automatic dependency installation is undocumented
- [Bug] userMessageBackground not rendering in fullscreen TUI mode on Windows
- get_mcp_status() drops MCP tool `title` — expose it in the mcp_status response
- --channels inbound message silently dropped: shown as pending draft, never committed as a turn, no trace in MCP receive log
- [BUG] iOS Simulator panel reports 'Xcode is installed but not selected' on Xcode 26.6 / macOS 26.5 despite correct xcode-select
- [Bug] Unexpected fallback to Claude 3 Opus when testing super admin functionalities
- Local peer-messaging inbox socket fails to bind for one of two identical sessions (one-way cross-session messaging)
- Support widget and bug reporter both dead: Intercom Messenger receives both user_hash and intercom_user_jwt
- claude-code-guide subagent cannot be invoked: fixed ~214k token overhead exceeds the 200k context limit
- macOS TCC permissions are bound to versions/<version> instead of the signed bundle, so Screen Recording and Accessibility must be re-granted after every update
- [FEATURE] Show device/session identifying info on Authorization Tokens page
- [BUG] Agents in any model is very slow when using Bedrock
- [BUG] Desktop app (Windows) crashes/hangs while using the built-in Claude Browser pane; one crash reset the app profile
- [Bug] Safeguards false positive downgrades model mid-task in authorized security workflows
Resolved issues (144)
- [DOCS] Npm update docs do not explain release-channel behavior for `claude update`
- [DOCS] MCP docs omit combined startup notification for MCP server and connector authentication
- [DOCS] `--fallback-model` docs still say it is ignored in interactive sessions
- [DOCS] `showTurnDuration` docs omit the background-agent/workflow waiting message and cumulative timing behavior
- [DOCS] Agent SDK docs omit `subagent_type: "claude"` worktree and output persistence behavior
- [DOCS] Plugin marketplace guide omits `skipLfs` option for git-based sources
- [DOCS] Fullscreen docs omit live "Thinking for Ns" timer and interrupt persistence
- [DOCS] Extended thinking docs omit collapsed summary rendering behavior
- [DOCS] `/usage` documentation omits large session files in limits breakdown
- [DOCS] Document server-side system prompt experiments and their opt-out controls
- [DOCS] /usage docs omit the current limits-usage breakdown categories
- [DOCS] `claude --bg --dangerously-skip-permissions` persistence across retire/wake is undocumented
- [DOCS] Worktree cleanup docs omit failed-removal safety behavior for gitignored and in-progress files
- [DOCS] PowerShell tool docs still describe opt-in rollout instead of default-enable behavior for Bedrock, Vertex, and Foundry on Windows
- [DOCS] `/plugin` marketplace browse pane projected context cost estimates are undocumented
- [DOCS] `/tui` docs omit that running background shells and subagents block renderer switching
- [DOCS] Interactive mode docs omit interrupted-prompt Up-arrow history behavior
- [DOCS] Extended thinking docs omit the amber spinner state during long thinking periods
- [DOCS] `/feedback` docs omit recent-session attachment options for cross-session issues
- [DOCS] Auto mode docs omit that `permissions.ask` rules still trigger manual approval prompts
- [DOCS] Plugin manager docs omit fullscreen tab and search navigation controls
- [DOCS] [Permissions] Agent(subagent) matching docs omit case- and separator-insensitive behavior
- [DOCS] Vertex AI workload identity federation docs missing `ANTHROPIC_WORKSPACE_ID` workspace-scoping guidance
- [DOCS] [Commands] `/context` output docs omit plugin names for plugin-sourced skills
- [DOCS] Gateway background session naming docs omit main-model fallback when no Haiku model is configured
- [DOCS] [Plugins] `claude plugin install` should document auto-refresh and retry behavior
- [DOCS] API-key authentication docs omit that it disables web-only Claude.ai features
- [DOCS] [MCP] `/mcp` reconnect picks up `.mcp.json` edits without restart and shows HTTP status/URL on failure
- [DOCS] Slash command docs omit mid-input autocomplete behavior
- [DOCS] Sub-agent docs omit `mcpServers` support for main-thread `--agent` sessions
- [DOCS] External editor docs missing last-response commented context option
- [DOCS] Remote Control docs omit web UI title persistence and precedence
- [DOCS] Prompt caching docs omit the startup warning shown when `DISABLE_PROMPT_CACHING*` disables caching
- [DOCS] Rewind command docs omit `/undo` alias
- [DOCS] `/model` docs omit the uncached full-history warning for mid-conversation switches
- [DOCS] Plan mode exit docs omit higher-permission restore behavior
- [DOCS] Queued-message docs omit attachment handling and focus-view visibility
- [DOCS] Session resumption docs missing /recap command and away-summary controls
- [DOCS] MCP output warning docs lack format-specific recovery recipes
- [DOCS] Plugin marketplace docs missing dependency auto-install behavior
- [DOCS] Worktree cleanup docs omit squash-merged PR cleanup behavior
- [DOCS] Environment variable docs are outdated for stalled stream recovery behavior
- claude.ai connector (Slack) shows Connected in VS Code extension but exposes zero tools to the model
- [DOCS] Loop command docs omit `/proactive` alias
- [DOCS] Sandboxing docs missing Linux subprocess isolation and `CLAUDE_CODE_SCRIPT_CAPS`
- [Bug] Anthropic API Content Policy: Overly Restrictive Domain Blocking for Multi-Disciplinary Research
- VS Code extension: remoteControlAtStartup: true is not honored — sessions never auto-start Remote Control
- remoteControlAtStartup is ignored by the VS Code extension
- Remote Control: remoteControlAtStartup never connects in VS Code/Cursor extension sessions (manual /rc works)
- remoteControlAtStartup not honored in VS Code / Cursor extension (works in desktop app + terminal CLI)
- [BUG] VS Code extension ignores `remoteControlAtStartup` — re-filing after #41036 and #53647 were auto-misrouted by dup-bot
- [BUG] Desktop app session list only shows the newest session after update (data intact on disk)
- [BUG] Cowork custom instructions: stale version served to new sessions
- Remote Control fails to enable on sessions with a large transcript ("Session creation failed — see debug log")
- Claude Desktop app import rewrites mtimes on all existing CLI session files
- [FEATURE] A hook that can transform the message array before each model request
- [DOCS] Remote Control docs omit session card repo, branch, and working directory details
- [DOCS] Agent SDK transcript viewer docs missing final token usage semantics
- [DOCS] Image input docs missing token-budget/compression details for pasted and attached images
- [DOCS] Sandbox settings missing `sandbox.network.allowMachLookup` documentation for macOS
- [DOCS] Syntax highlighting docs missing Cedar policy file support
- [DOCS] Subagents docs missing `/agents` running count indicator
- [BUG] ant auth login switches Claude Code to API billing
- [DOCS] Sandboxing network-access prompt flow is outdated for auto mode and bypassPermissions
- [DOCS] Plugin docs missing `bin/` executables and Bash bare-command usage
- [DOCS] PowerShell tool docs omit dangerous-command safety behavior
- [DOCS] Remote Control docs omit session-list "Requires Action" status for permission prompts
- [DOCS] Interactive mode docs missing queued-prompt editing and slash-command interaction
- [DOCS] MCP policy docs do not explain `deniedMcpServers` blocks Claude.ai connectors
- [DOCS] Plugin enable/disable docs omit scope auto-detection when `--scope` is omitted
- [DOCS] Scheduled tasks docs missing transcript timestamp markers for `/loop`/`CronCreate` firings
- [DOCS] `/stats` command docs missing Ctrl+S screenshot shortcut
- [DOCS] VS Code docs missing rate limit warning banner with usage percentage and reset time
- [DOCS] Background bash command docs omit ~45-second stuck-prompt notification
- [DOCS] VS Code extension docs missing "Not responding" red spinner indicator after 60-second backend timeout
- [DOCS] `cleanupPeriodDays` setting docs omit tool result files from cleanup scope
- [DOCS] LLM gateway docs missing `x-client-request-id` header for timeout debugging
- [DOCS] MCP docs missing non-blocking startup behavior for `claude "prompt"`
- [DOCS] Document Linux `XDG_DATA_HOME` support for `claude-cli://` protocol handler registration
- Code tab lost its "Recents" group after update (Home tab still has Recents)
- [BUG] Persistent "Unable to connect to API (ECONNRESET)" — survives network change, reinstall, and reboot (macOS, v2.1.212, Claude Pro)
- test
- [BUG] Workflow tool's per-agent model override is silently ignored for subagents when running headless (claude -p)
- [BUG] /sandbox Config shows "Network Restrictions" allowlist that doesn't restrict when autoAllowBashIfSandboxed is true
- [BUG] VSCode Extension: /model command output inconsistent across sessions
- Claude Code conversation history doesn't sync across computers on the same account
- [BUG] Paid Max acct routed to /onboarding after authentication
- [BUG] Windows: pop-out session windows have no min/max/close buttons and cannot be dragged - WS_SYSMENU missing after setTitleBarOverlay fails
- [BUG] Windows: New windows have no draggable title bar area, making them impossible to move
- [BUG] Windows: floating/pinned conversation windows cannot be moved or closed with the mouse — no window controls, setTitleBarOverlay always fails
- Detached session window cannot be maximized or fullscreened (Windows)
- [BUG] [Desktop/Windows] Chat opened in a New Window (pop-out) has no title bar - can't drag or resize; regression after MSIX update to 1.25927.0
- [BUG] Cowork: folders attached in Project → Context are not mounted in new chat sessions
- [BUG] Claude Desktop's built-in terminal can never get Apple Events (Automation) permission
- [BUG] Desktop: PR links always open in the internal Browser pane while every other external link opens in the default browser
- Assistant text preceding an AskUserQuestion tool call is not shown to the user
- .claude/worktrees/ directories are silently recreated after deletion, even with Desktop closed and no owning process found
- [BUG] Fullscreen TUI: failed Bash tool calls drop the entire ⎿ result row (stdout+stderr present in JSONL and tool_result; success path OK)
- computer-use MCP: browsers reported "(not installed)" on macOS when Spotlight is disabled, despite LaunchServices having them registered
- [DOCS] Document plugin MCP server suppression when duplicating an org-managed connector
- [DOCS] Interrupting a prompt before any response should be documented: input is automatically restored for editing and resubmission
- [DOCS] Document the non-streaming API fallback mechanism, including its token cap and timeout behavior
- [DOCS] Claude.ai MCP connectors (Slack, Gmail, etc.) availability in `--print` / `-p` mode is undocumented
- [DOCS] Document diff timeout behavior for large files with few common lines
- [DOCS] Transcript search shortcut (`/`) and match navigation (`n`/`N`) not documented
- [DOCS] `--mcp-config` CLI flag now subject to `allowedMcpServers`/`deniedMcpServers` managed policy — docs don't mention this enforcement applies to CLI-loaded servers
- [BUG] MCP server args containing `~` (tilde) cause ENOENT — tilde not expanded in stdio server arguments
- [DOCS] Model configuration page missing documentation of automatic fallback notification behavior
- [DOCS] Remote Control docs outdated — bridge session reconnects in seconds, not up to 10 minutes
- [DOCS] Permissions and settings docs missing bash auto-approval allowlist enumeration
- [DOCS] VS Code docs missing permission-picker behavior when `disableBypassPermissionsMode` is set
- [DOCS] Interactive mode docs missing numeric keypad option selection behavior for interview prompts
- [DOCS] MCP/WebFetch docs missing binary output file-handling behavior introduced in v2.1.69
- [DOCS] CLI reference lacks skill-loading semantics for `--setting-sources user`
- [DOCS] Plan feedback docs do not explain multiline input behavior when rejecting plans
- [DOCS] Agent SDK TodoWrite docs still require `activeForm` after v2.1.69 made it optional for task creation
- [DOCS] Hooks reference lists incomplete set of matchable tool names for PreToolUse and PermissionRequest
- [DOCS] `awsAuthRefresh` and `awsCredentialExport` timeout behavior is undocumented
- [BUG] SessionStart hook output silently truncated at 10KB — model never sees the missing content
- [DOCS] Plugin MCP deduplication docs imply same-command servers are always duplicates
- [DOCS] [additionalDirectories] Read/Edit tools denied on mapped network drives — fixed in v2.1.133
- [DOCS] Interactive mode docs missing native terminal IME composition and screen-reader caret tracking
- [DOCS] Plugin docs missing autocomplete for `/plugin` arguments: subcommands, installed plugin names, and marketplace plugins
- [FEATURE] BeforeModel and AfterModel Hooks for LLM Request/Response Interception
- [DOCS] `/web-setup` docs omit warning before replacing an existing GitHub App connection
- [DOCS] WebFetch docs omit HTML preprocessing and style/script stripping behavior
- Streamable HTTP client mis-correlates JSON-RPC error body of 405 GET (SSE stream) response to pending tools/list, breaking stateless servers (Linear MCP)
- [Feature Request] Rate Claude Session prompt should use distinct numbering separate from solution options
- [DOCS] MCP docs omit remote tool idle timeout and `CLAUDE_CODE_MCP_TOOL_IDLE_TIMEOUT`
- [BUG] HTTP MCP client drops Mcp-Session-Id header — tools/list fails on session-aware servers (mcp-grafana, mcp-go) despite ✓ Connected
- [FEATURE] Worktree session names should show task content, not hostname-based branch names
- [BUG] EnterWorktree tool does not process `.worktreeinclude` — gitignored files not copied, unlike `claude --worktree`
- RFC: Async / event-driven communication as a first-class capability for Claude Code agents
- [Bug] Tool invocation XML namespace dropped causing malformed tool calls in Opus 4.8
- Code tab: add bulk-delete for sessions, and don't persist sessions that never submitted a prompt
- iOS app crashes immediately when tapping a session in the Code tab
- [BUG] Claude Code Analytics not updated since May 12
- Stop hook exit-0 output not reliably surfaced in hookAdditionalContext
- [Bug] Safety check flagging something that is clearly not remotely at risk of violating usage terms.
- [BUG] iOS app 1.260618.0 hard-crashes opening a Remote Control session — main-thread - stack overflow in Swift KeyPath metadata
- [Bug] Content filtering policy overly restrictive, blocking legitimate requests
- [BUG] "Unsupported File Type" Eror panel can not be dismissed
- [BUG] Claude Desktop Electron main process grows unbounded (~3.7 GB/h) to 47 GB, triggering system-wide jetsam on macOS
- [BUG] Artifact tool not loaded in interactive Linux CLI session — same account works on macOS CLI (gate allow_cobalt_plinth?)