Digest / September 2, 2026
Claude Code Issues Digest — September 2, 2026
300 new issues filed ·
503 resolved in anthropics/claude-code (UTC day).
New issues (300)
- [Bug] False positive safety filter: legitimate database code flagged as biohazard content
- [FEATURE] Setting to suppress update notification without disabling auto-update
- [BUG] Windows installer cannot recover when the MSIX install volume is physically removed — orphaned package + stale packaged service block AddPackage permanently
- [BUG] Can not /login via SSH - URL is not displayed nor is it copy-able
- [Bug] EDR Log Analysis Task Produces False Positives
- [Bug] CLI resolves wrong Railway project from current directory without warning
- : Safety classifier blocks legitimate defensive security work on own infrastructure, broadens scope after triggering
- Cross-session peer messages collapse to a one-line preview since 2.1.247
- [BUG]
- [Bug] Vague blocking error prevents local development workflow
- [Feature Request] Agent duo companion mode for collaborative task execution
- [BUG] Claude Desktop (Windows/MSIX) causes recurring kernel BSODs - 4-month evidence incl. a 7-week natural experiment - same family as #30137 / #69359 / #80912
- [Bug] Repeated identical errors due to instruction parsing failure
- Accessibility: distinguish subagent permission prompts without relying on color
- [Bug] Model silently downgraded from Opus 3.5 to Opus 4.8 mid-session without error notification
- destinationgod@gmail.com - unable to include claude extension
- [BUG] Local scheduled tasks silently hang mid-run and block later scheduled fires
- Cross-session messages held for approval are indistinguishable from delivered ones at both sender and recipient
- macOS: desktop-launched Claude Code is silently denied Local Network access and never gets the permission prompt (CLI bundle is LSBackgroundOnly)
- [BUG] Chat Title from VS code claude code does not sync to desktop claude code.
- [Bug] Unexpected model switch to Opus 4.8 during testing
- [BUG]
- [BUG] Artifacts viewer window: View > Zoom In/Out and Ctrl+/Ctrl- zoom the main Claude Code window instead of the artifact
- [Bug] Legitimate code incorrectly flagged as malicious or unsafe
- [Bug] Session cost tracking unclear when code review stalled mid-task
- [Bug] Content Safety Filter Incorrectly Downgrades Model for Legitimate Payment Security Analysis
- [BUG] Desktop app's bundled Claude Code crashes on macOS 12 (dyld: _DNSServiceGetAddrInfoEx) while the standalone CLI runs fine
- [Bug] Safeguard blocks access to internal system management tasks
- Fabricated user messages injected into model context with literal "user" role marker prefix
- [BUG] /remote-control stopped working unexpectedly yesterday morning
- Context ring no longer warns before the window limit — session died at 'Prompt is too long' with no prior warning
- Agent silently ran a 10x more expensive model than instructed and reported the cheaper one
- Expose scheduled task concurrent session limit as a user-configurable setting
- [Bug] Ctrl+Click on URLs opens duplicate tabs instead of single tab
- [Feature Request] Add allowlist/exception mechanism for legitimate biotech research use cases
- [Bug] Anthropic API Error: False positive content filter block on benign input "Hi"
- [Bug] Fable 5+ consistently ignores defined skills and creates incorrect memory workarounds
- Same session appears under three independently-generated names, breaking identification and cross-session addressing
- [BUG] Artifact fails to load intermittently — "Connection reset" error
- [Cowork/claude.ai] Cannot add anthropics/claude-plugins-community as a marketplace; community plugins (e.g. eli5) missing from the directory
- [BUG] Artifact tool loads ~12k tokens of schema into every claude.ai-backed session, eagerly, with no way to opt out except a deny rule
- Assistant emitted a fabricated user message and acted on it as authorization (ran git commit)
- [Bug] False positive security flag on internal database queries
- [Bug] Agent fails to maintain context in multi-step coding workflows
- [Feature Request] Add session token refund mechanism for failed/unproductive sessions
- Desktop app scheduled-task runs never exit their claude session process; they accumulate at ~5% CPU each until the Mac saturates
- Display per-message timestamps in chat/session view
- [BUG] Startup cleanup deletes live sibling sessions' Bash task output under the shared project temp root
- [Bug] Insufficient information provided for bug report
- Ctrl+Alt+K (add file to context) stops working after /clear + opening a new terminal in IntelliJ Description:
- Worktree pool assigns relaunched sessions to the wrong worktree (95% of sessions measured), and can discard uncommitted work
- Desktop app (Windows): no way for the agent to read its own 5-hour usage
- [Bug] Session flagged by content filter despite model confidence
- Chained approve && merge defeats interrupts; ambiguous follow-ups resolved toward production-bound actions
- Windows: app state silently lost when %APPDATA%\Claude is a junction — data directory not configurable
- [FEATURE] Make the auto-mode permission classifier timeout configurable
- [BUG] "Your account has been suspended". Claude Max20 account banned one day after a $249 payment — no reason, warning, or refund
- [Feature Request] Add command-line flag to set agent color when scheduling agents
- [BUG] Individual sessions degrade 10-30x in latency while concurrent sessions on the same project, version and model stay normal
- [BUG] MCP HTTP: first turn blocks for MCP_TIMEOUT-5000ms awaiting subscriptions/listen after a successful server/discover
- docs: subagent frontmatter `effort` field is implemented but undocumented
- [BUG] claude.ai web: private personal repo not accessible via "Add from GitHub" despite Claude GitHub App installed with access granted (backend index/sync issue)
- [BUG] Mid-input slash completion never matches plugin-namespaced skills (plugin:skill), while the position-0 dropdown does
- [BUG]
- [BUG] StopFailure fires for internal helper queries and subagent failures on the parent session; under low-priority mode every excluded helper 429 reports a failed turn that never happened
- In-chat search
- [Bug] Increasing latency on simple tasks over time
- `--permission-mode dontAsk`: `cp`/`mv` refuse any option token (including bare `--`) while the flagless form runs; `rm` is unaffected
- [Bug] Infinite loop in code execution
- Opus 5: explicit user directives overridden by inferred defaults on a prose-writing task; attention misallocated to unrequested work; degrades after correction
- Claude Code has no way to read/import claude.ai Project (Cowork) knowledge files
- [BUG] NODE_OPTIONS=--use-system-ca replaces instead of augments the bundled CA store — confirmed regression #72066 was auto-closed as not_planned
- [Bug] Android notifications not working on macOS with iTerm2
- [BUG] Dispatch missing from Claude Desktop on Windows — visible on mobile (same account), pairing always fails (related: #82431)
- [BUG] Buttons not working in Desktop Claude code
- [Bug] Custom validator triggers false positive spam detection
- [BUG] VS Code extension: deleting a session is instant and permanent - no confirmation, no undo, transcript unlinked
- test
- [BUG] Desktop (Linux ARM64): session registry saves fail with EINVAL — sessions vanish from sidebar on restart
- You've hit your session limit · resets 3:30pm (Asia/Kolkata)
- [Bug] Unexpected Model Change During Execution
- [BUG] VS Code extension logs out after idle/overnight, requires full re-login despite being logged in on claude.ai
- [Bug] Assistant inefficiently debugging CSS issues, consuming excessive tokens on symptom fixes instead of root cause analysis
- [BUG] Desktop (Windows): clicking a file/preview link in a detached session window does nothing — the preview pane never opens (works in the main window)
- Fable guardrail fire on completely unrelated project
- Cowork/synced plugin packs can't be listed, inspected, or disabled; claude plugin details can't resolve them
- Log in opens to onboarding create your account
- Desktop: slash command menu does not open when input box already has text
- [BUG] Desktop: staged worktree checkout runs before .gitattributes is in the index, so .claude/** and CLAUDE.md get CRLF on Windows (core.autocrlf=true)
- [FEATURE] Bind effort level to permission mode (xhigh in plan mode, default elsewhere)
- [FEATURE] Official WhatsApp MCP server/plugin that works with personal accounts
- [BUG] Permission mode selector not shown in the iOS composer during an active Code session
- [BUG] MacOS kills Claude and says "Malicious Script Blocked"
- [BUG] all new sessions created with remote session enabled despite never making this setting
- [FEATURE] Let in-flight subagents resume from their own checkpoint after a usage-limit interruption, instead of restarting from scratch
- Monitor: log-follow watch gives no signal when the log's writer dies; request a liveness/exit event
- I appreciate your interest, but this isn't a bug report or feature request for Claude Code. This appears to be a job inquiry. If you have a genuine bug to report or feature to suggest for Claude Code, please provide details about: - What problem you encou
- [Bug] Agent files handoffs into the repo of a recently-read doc, not the repo that owns the work (salience anchoring; skips its own placement rule, then justifies post-hoc)
- [Bug] Safeguards trigger blocking legitimate user requests
- [BUG] Desktop: forking a session duplicates its artifact watch, so a publish from the original wakes the fork into an unrequested turn
- [BUG] claude-in-chrome: get_page_text/screenshot/read_page all time out on LinkedIn Jobs pages ("still loading" / document_idle never resolves)
- I appreciate your interest, but this isn't a bug report or feature request for Claude Code. This appears to be a job inquiry. If you have a genuine bug to report or feature to suggest for Claude Code, please provide details about: - What problem you encou
- [BUG] claude generate a answer start with "user" and think it's my prompt.
- [BUG] Assistant ends the turn announcing work it never starts — silent no-op, user must poll to discover it
- [FEATURE] allow to disable remote control completely
- Sync conversation history per repository, across machines
- [BUG] German translation of "Running command" is not "Läuft Befehl" but rather "Befehl wird ausgeführt" or "Führe Befehl aus"
- [BUG] Fable 5.1 sessions replace the assistant's interim text with API summaries ("summarized" lines), undocumented and with no visible off switch
- Google Drive connector `create_file` always produces Letter-sized Google Docs, ignoring account default and content-level page-size hints
- [DOCS] Desktop transcript retention: the "while they are in the app" condition on the sweep exemption is not documented
- [VSCode] Session Manager lists a worktree-relocated session but opens a blank tab for it (2.1.258); resume of a deleted-worktree session leaves the transcript stranded
- [BUG] Desktop SSH remote: session init phase blocks on a ~600000ms timeout before succeeding (8/13 sessions affected)
- [BUG] Current session" usage shows high % on a fresh session despite ~12h of prior inactivity; confirmed at account level (claude.ai), not just local CLI display
- [FEATURE] Honor hookSpecificOutput.sessionTitle from PostToolUse/Stop hooks (mid-turn session rename)
- [SOLVED] Claude in Chrome "not connected" on Windows: stale oauth:tokenCache causes a 401 loop — fix inside
- stdio MCP servers stop responding after system sleep/wake; require full session restart to recover
- [Bug] Rate limit consumed prematurely - 5 hour window exhausted in 30 minutes
- [FEATURE] Make git worktree pooling opt-out: one worktree per session, removed only when the session is archived
- [FEATURE] Field notes after a year of home-grown memory on Claude Code: two bugs, four requests, and a thank-you
- [Bug] Anthropic API Error: Messages consistently flagged across all models
- [BUG] Numbers are changed between writing and sending to Claude Code.
- You've hit your session limit · resets 6pm (Europe/Warsaw)
- [MODEL] Fable 5.1 defaults to using Bash on windows no matter how many times I correct it.
- [MODEL] Assistant fabricates user turns in its own output, then acts on them as if the user sent them
- [BUG] blockReadsOutsideWorkingDirectories prompts for allowlisted gh commands: flag values misidentified as runtime-computed paths
- [BUG] Auto-update at launch runs for bare `claude` but is skipped when launched with `--effort max`; the flagged session stays on the old version
- [BUG] Internal browser crashes when accessing Windows Credential Manager during authentication
- [Bug] MSIX auto-update leaves package in unlaunchable state: 0x80070020 container creation failure
- [Bug] Code review on large branches hits session timeout before completion
- [BUG] Scroll does not work anymore in the terminal for the Claude code output
- Claude in Chrome extension not connecting (native host registered correctly)
- [Bug] Session limit reached prematurely; automatic session resumption after reset
- [Bug] Session timeout reached prematurely after Fable 5.1 upgrade
- Fable 5.1 unreachable in Claude Code despite unused plan-included Fable allocation; credit-block dialog shows "not included in plan" copy with no affirmative action
- I appreciate you reaching out, but I'm not able to generate a GitHub issue title from this report because: 1. The text appears to contain typos/unclear phrasing ("false flag", "dovlepoing", "resturant") 2. There's no clear technical problem or error messa
- [BUG] Claude Code process exited with code 134
- [Bug] Anthropic API Error: Persistent API errors during code execution
- [BUG] Grep/Glob removal on native macOS/Linux is a breaking API change, not a docs gap (re: #51781)
- Usage-limit auto-continue doesn't resume workflows or subagents — they all fail
- [BUG] Critical: Claude Code Hit 5 hour limit after only 2 minutes and 21% Fable 5.1 usage
- [BUG] Site permissions ("Allow all websites") ignored by the built-in browser in Claude Code Desktop
- Title: Cowork session hangs silently after 2nd prompt — no error in logs, VM stays Running Environment:
- [FEATURE] Split view as a primary new session experience
- [BUG] `Bash` tool executes the user's login shell (e.g. /bin/zsh) and should thus be called "shell" not "bash"; LLM trips over bash-only idioms
- Request ID: req_011Ceef8QLMr3omGcLC6dLf3...
- [BUG] Claude Code IDE's WSL2 folder picker mis-resolves symlinks that Cursor/VS Code Remote-WSL handle correctly (architecture-specific, not a WSL limitation)
- [DOCS] Claude Code iOS
- [BUG] Established connections to the API go silent; client waits 180s+ before retrying (retry then succeeds in ~1.5s)
- [BUG] Windows: a subagent's background-task completion notification is queued but has no idle-wake consumer — the subagent hangs until something else gives it a turn (refs #78338, #21048, #29163)
- Windows MSIX: GPU process killed by Code Integrity loading bundled vk_swiftshader.dll; app terminates and cannot relaunch
- I appreciate you reaching out, but I need to clarify my role: I'm Claude Code, Anthropic's CLI tool for code generation and development tasks. The message you've sent appears to be in Indonesian and discusses event judging procedures, spreadsheets, and PD
- Auto mode ran a destructive GitHub ProjectV2 mutation (iteration-list replace) unflagged — wiped 12 completed sprints; classifier then blocked the identical restore call
- [Feature Request] Reduce response verbosity and prioritize direct solutions over caveats
- DISABLE_FEEDBACK_COMMAND=1 not honored in the VS Code extension — /feedback and /bug remain listed and fully functional
- [BUG] VS Code extension: showMessageTimestamps, timeFormat and timeZone are documented in the schema but never read
- [BUG] Windows Desktop application sidebar groups can't be set to custom
- Desktop app: input typed into a Windows session executes in an unrelated macOS session (cross-machine session mix-up)
- Sandboxed Bash leaves per-invocation writable-overlay tmp directories behind; never cleaned up, fills disk at several GB/hour
- In a shared multi-agent working tree, `git commit --amend` doesn't re-check HEAD, so it can silently rewrite a concurrent teammate's commit instead of the agent's own
- [BUG] Warm prompt cache fully re-written (cache_read collapses to system+tools floor) seconds after ToolSearch / Skill / tool_result on v2.1.258 — 4 events with request IDs; plus pre-2.1.257 screenshot size-cap data
- [FEATURE] Weekly-limit boost ends 13 Sep: four months in, it's the baseline, not a bonus
- [Bug] Model switching from Fable Sonnet to Opus mid-conversation degrades output quality
- Auto-mode safety classifier intermittently unavailable for Write/Bash/Edit, blocking work for extended periods
- Emit --append-system-prompt as its own system block so a changed append does not invalidate the preset
- [FEATURE] Signal unsent draft input in the terminal tab title (OSC), so multiplexed sessions show which tab is waiting on you
- settings.json is written with an unlocked, non-atomic read-modify-write: concurrent sessions tear it ("Settings file failed to parse") and silently drop each other's keys (2.1.258, with repro)
- [FEATURE] Design Mode: click-to-select an element in the Browser pane and attach it to the prompt
- [Bug] Unintended model switching to Opus overrides user model preference
- Background-task cleanup leaves orphaned find.exe processes on Windows (Git Bash/MSYS)
- [BUG] System incorrectly downgraded Max Plan to Free after requesting a refund for unapplied usage credits
- [FEATURE] Desktop app: make a project a saved multi-folder config with its own new-session plus
- [BUG] Windows: Claude Code CLI corrupts stdio MCP password arguments when launching npx through cmd.exe
- Scheduled task: scheduler skips runs / reports success (lastRunAt updates) with no session or output produced
- [BUG] Desktop app: crash/interrupted update corrupts session sidebar — projects reset to "Other", titles/order lost, sessions appear merged (possible same root cause as #76430)
- [BUG] API Error: Claude Code is unable to respond to this request, which appears to violate our Usage Policy (https://www.anthropic.com/legal/aup).
- [BUG] `claude --resume` exits at startup under tmux — probe replies read as input
- File attachments arrive with empty content in Desktop chat (Windows) — both drag-and-drop and file picker affected
- claude-fable-5-1 appears to over-provision parallel agents in self-authored Workflow scripts vs claude-fable-5 — 25 agents vs 15-limit default, ~10x cost on an equivalent step
- [Bug] Claude Code auto-approves plan mode without human intervention
- [Bug] Auto-approval of plans and automatic mode activation without user confirmation
- Auto-continue after usage-limit reset (2.1.234+) doesn't fire in a detached tmux session
- VS Code extension: Escape interrupts running task in native UI, with no setting or keybinding to disable it
- PII filter incorrectly sanitizes Figma node IDs in MCP tool responses
- [BUG] Desktop SSH: quitting the app opens N simultaneous ssh connections (N = sessions on the host, archived included) instead of using the existing control socket
- [BUG] ECONNRESET, four times in ~20 minutes
- [BUG] Workflow approval rejects benign CRLF for all file-backed invocations (name + scriptPath) on Windows
- Proposal: durable, cross-runtime messaging protocol for Agent Teams (design + validated implementation)
- [BUG] Remote HTTP MCP server never re-probed after a transient network drop; misreported as UNKNOWN_CERTIFICATE_VERIFICATION_ERROR
- [Bug] Display switching from Vision Pro virtual display causes crash or error on return to Studio Display
- [Feature Request] Add Google Drive integration for file uploads
- [Bug] Session-wide permission consent not persisting in Chrome extension
- Claude-Session trailer still appended to commits and PR bodies with includeCoAuthoredBy: false (regression of #66504)
- Browser pane can't grant local-network/loopback permission, so Duo Desktop device-health check fails (Duo-protected SSO sites unusable)
- [FEATURE] Feature request: lazy/on-demand OAuth re-auth for MCP servers (4 related issues closed/locked)
- Fable 5.1 in Claude Code ignores a standing "delegate, don't code" instruction and burns premium usage on grunt work
- [BUG] dyld Symbol not found (_DNSServiceGetAddrInfoEx) crash on macOS 12.7.6 — claude-code 2.1.255
- [BUG] Paste misplacement: re-pasting text identical to the last collapsed paste expands that old placeholder and moves the cursor there, instead of pasting at the cursor
- Claude connection not working
- [Bug] Claude Code: Inconsistent command execution and response accuracy across sessions
- [Feature Request] Cowork projects: GitHub-synced files aren't individually browsable (only RAG-searchable)
- [BUG] macOS sandbox: System V shared-memory and sysctl calls are refused with EPERM whenever two sandboxed Bash commands run at the same time — and a refused shmget/shmctl still leaves the segment allocated
- [Bug] Anthropic API Error: Overly aggressive safeguards blocking legitimate security code review
- [Feature Request] Add in-game chat button UI component
- Worktree provisioning writes the git exclude entry to a literal --git-common-dir/ directory in the MAIN checkout instead of .git/info/exclude
- [Bug] Terminal display corruption: text overlaps after long Bash commands
- [BUG] VS Code extension: repeated Remote Control (re)registration mints a fresh session codename each time — Local tab fills with near-simultaneous duplicate rows
- Cowork: mcp__workspace__bash and web_fetch permanently denied — host passes Bash/WebFetch in --disallowedTools, which aliases onto the workspace tools
- Cowork projects migrate from personal account to Team account after sign-out/sign-in on the same Windows install; archive is non-functional
- [Opus 5] Doesn't follow the instruction at the user level
- [Feature Request] Rename method to Authorities
- [BUG] Cowork local sessions: all mcp__workspace__* tools denied without any prompt after SDK 2.1.258 (Windows)
- [Bug] Cyber Verification Program status not honored in API requests, safeguard blocks legitimate learning plan requests
- [BUG] Cowork (Windows): mcp__workspace__bash and web_fetch instantly denied in every session, no dialog — related to #57101 / #74324
- Allow 'cd + git' commands without security prompt when inside additionalDirectories
- [BUG] macOS: main window left at Floating level (layer=3) after Computer Use side panel restores
- [BUG] XDG Path improvement
- Windows: refresh-token rotation not persisted to .credentials.json; failed refresh blanks the store, locking out all fresh processes (2.1.220)
- Auto-promote a session to the top of the pinned list after dwelling in it (~15s)
- [FEATURE] Bring the CLI session recap (/recap and the away summary) to the desktop app
- PreToolUse hook on Write/Edit/MultiEdit/NotebookEdit is not enforced (PostToolUse on same matcher fires reliably)
- [BUG] Windows desktop 2.1.258: 'Split view' missing from the 'Open in' submenu (shows New window / Explorer / Cloud)
- [Feature Request] Remove unnecessary prompt flagging on authenticated dashboard
- [Bug] Anthropic API Error: Safeguards flagging legitimate cybersecurity task as policy violation
- Windows desktop app: idle warm sessions never release MCP child processes (1 to 1,182 children / 33 GB RSS in one day, ends in GPU process crash)
- [Feature Request] Implement semantic content detection for safety moderation instead of keyword-based filtering
- [BUG] Background task-notification bodies silently truncated at a hardcoded 100,000 chars — TASK_MAX_OUTPUT_LENGTH does not apply and no saved-output path is included
- Windows: `CLAUDE_CODE_SHELL_PREFIX` is used as the executable for stdio MCP server spawns, so a bash-script prefix kills the first connection ("The system cannot find the path specified.")
- Cowork: mcp__workspace__bash and web_fetch denied in all sessions after desktop 1.44121.1 update (Windows)
- [BUG] Prompt autocomplete moved above the input and shows only 2-3 matches, with no way to configure it
- [Bug] False positive security detection flags legitimate CVE research activities
- [Bug] Anthropic API Error: Content policy flagged legitimate cybersecurity work despite cyberverification approval
- VS Code extension 2.1.258 hangs indefinitely (stuck spinner) after update; reverting to 2.1.257 fixes it
- Desktop app (Windows): hard crash right after the clipboard-change warning during dictation (Wispr Flow)
- Allow browsing the file tree before starting a chat (desktop app)
- [BUG] Desktop Code tab: resumed session drops claude.ai connector tool schemas, array/number arguments sent as strings (-32602)
- I'm ready to help, but I don't see a bug report or error message in your message. Could you please provide: 1. The error message or unexpected behavior you're experiencing 2. What you were trying to do when it occurred 3. Any relevant output or logs from
- LSBackgroundOnly ignored: background/Cowork sessions each spawn a visible Dock icon
- [BUG] Remote Control sessions require PC-side input to reconnect after restart; phone-side reconnection fails
- I don't have enough information to generate a GitHub issue title from just a Request ID. To create an appropriate title, I would need details about: - What the actual bug or issue is - What error message occurred (if applicable) - What the expected vs act
- [BUG] Cowork (Windows): mcp__workspace__bash denied in every session after updating to 1.44121.1.0 — worked on 1.40609.1.0
- [BUG] Opus 5 never performs context discovery — project roadmap, CLAUDE.md rules and hooks all present but none load before acting (Codex/Fable unaffected in same repo)
- [Bug] Agent prefers bash heredoc over file write tool due to classifier misconfiguration
- [Bug] Agent loop creates duplicate tasks with excessive API usage
- [Bug] Agentic loop: Claude creates unnecessary issues and tickets recursively without convergence
- [BUG] Claude Code wiped an external USB drive after treating a clarifying question as authorization
- [BUG] Shell snapshot should not pull in user aliases by default
- Autonomous /housecarl-style loop: Stop-hook goal-condition check re-fires identically forever, ignoring explicit user stand-down
- [Bug] Security token leaked in session data
- [Bug] Overly aggressive content filter blocking legitimate security hardening discussion
- [BUG] Cowork regression: existing Scheduled Tasks + Live Artifacts broke after Aug 29 migration (macOS)
- [BUG] Cowork (Windows): every mcp__workspace__bash call denied "by rule" since Desktop 1.44121.1 / Claude Code 2.1.258 — --disallowedTools Bash now propagates to the aliased Cowork tool
- Accessibility (Windows desktop app): notification-area icon has an empty tooltip, screen readers announce an unlabeled button
- CCD: sessions auto-archive on PR merge without user action
- [Bug] Code quality analysis not functioning properly
- [Bug] Claude refuses to modify configuration settings despite explicit user authorization
- [BUG] Cowork: publishing an artifact silently stops updating its local copy in ~/Claude/Artifacts, and the directory mtime hides it
- [FEATURE] Cowork can't read the user's own published artifacts — no authenticated read path, unlike Claude Code
- [BUG] Claude Code Rovo Connector not working
- [BUG] Desktop app: plan file links can never be opened — ~/.claude/plans is never added to the File Pane's allowed roots
- Hook output over the 10,000-character cap is silently collapsed to a 2 KB preview with no user-visible warning (ref #44086)
- [Bug] Anthropic API Error: Safeguards falsely flagging legitimate game library code
- [Bug] Anthropic API Error: Unexpected Request Refusal on Valid Input
- [FEATURE] allow --disable-tips to hide any UI text that starts with "Tip:"
- Windows: isolation-worktree safety check rejects valid worktree due to case-sensitive drive-letter comparison
- [FEATURE] Cowork Dispatch: add `model` parameter to `start_task` so the orchestrator can route child tasks by tier
- Claude in Chrome: multiple connected browsers force an unanswerable per-session picker; need per-project browser profile bindings
- Desktop app (Code tab) crashes when Browser pane hits Cloudflare challenge while VPN is active
- [BUG] Cowork (Windows, host-loop): every mcp__workspace__bash call denied since Claude Desktop 1.44121.1 / Claude Code 2.1.258 — the app's own --disallowedTools Bash now matches the workspace shell via toolAliases
- Max 20x plan drained in ~22 hours after Fable 5.1; 5.1 emits 2x the output per turn of Fable 5
- [FEATURE] Batch tool use approval as a call-scoped token
- Desktop-bundled Claude Code (2.1.255) crashes on macOS 12 Monterey with dyld Symbol not found: _DNSServiceGetAddrInfoEx — worked until yesterday
- [BUG] Desktop app grows to 9 GB and dies; archiving 45% of the local session store fixes it
- [BUG] Remote Control: shutdown archives a constructed `session_` ID and books a clean SIGTERM exit as `failed`, so `--continue` finds no session
- Worktree-origin resolution keeps spawning git against deleted directories, breaking spawn_task chip launches (~10% failure rate)
- [Bug] Excessive permission prompts for allowed directories in auto mode
- [Bug] Non-biomedical task incorrectly classified as bio task
- "Add marketplace" stuck on "already added" after marketplace was removed
- Bridge/queue-delivered user turns are parented to a stale leaf, forking the transcript and silently stranding hours of work (3/3 forks in one session, distinct from #89820)
- [BUG] Startup FileIndex blocks the event loop for ~41s when cwd is a large non-git directory (home dir, ~65k files)
- [BUG] Recursive FileChanged watchPaths registration over a large NFS directory blocks the main thread at startup (D state, rpc_wait_bit_killable)
- Sidebar session groups/pins stop rendering after restart (backend data intact, confirmed via list_sessions)
- [Cowork] Windows host-loop: `mcp__workspace__bash` denied in every session after Desktop 1.44121.1 / CLI 2.1.258
- [Bug] Anthropic API Error: Cyber guardrails incorrectly flagged for legitimate web performance work
- [Bug] Text selection and handling is broken: auto-copy toggle doesn't restore Command-C copy functionality and command-double-clicking links doesn't work
- [Bug] Slow deployment script execution with repeated manual configuration steps
- [BUG] # [BUG] Windows (MSIX) 1.44121.2.0: app writes no logs to any location and Crashpad produces no reports — crashes and launch failures are undiagnosable from the user side
- Remote MCP OAuth: shared user-scope refresh token breaks against IdPs with single-use refresh-token rotation (ODC)
- Scheduled-task CLI process does not exit after unattended run completes
- plugin eval (macOS, ≥2.1.251): sandbox denies every Bash write — /private/tmp denyWrite covers the /tmp/e-* sandbox root
- Status line shows an opaque, undismissable "MR !<iid>" chip after creating a merge request; persists across restarts and leaks into other sessions
- [Cowork][Windows] CoworkVMService idle-shuts down the VM, then the app spins on the missing named pipe forever and never restarts the service
- autoUpdates: false in ~/.claude.json is ignored on native installs; updated twice in two days, once mid-session
- I don't see any image attached to your message. Could you please share the bug report or image you'd like me to create a GitHub issue title for? Once you provide the details, I'll generate a concise, technical issue title following the format you specified
- [BUG] Windows: two Bash tool calls never return and never time out — main session deadlocks for 53 min; agent completions and user input queue behind the dead turn
- [Desktop] Sort a project's Recents list by last message (the displayed timestamp column)
- [BUG] Bash cd-compound-read guard prompts on absolute cd targets whenever a Read() deny rule exists (Windows Git Bash, 2.1.257-2.1.259)
- [Bug] Fable Safety Filter False Positive on Public Open Source Code
- Left arrow never opens agent view in some sessions (persistent for the whole session, retrying doesn't help)
- [Bug] UI layout misalignment after toggling agents view on Windows 11
- [BUG] Microsoft 365 connector never requests Calendars.ReadWrite / MailboxSettings.ReadWrite — calendar write tools fail with 403 even when connected
Resolved issues (503)
- [Bug] False positive safety filter: legitimate database code flagged as biohazard content
- Background Tasks panel shows a CronCreate one-shot job as a recurring 'Loop' after CronList confirms it's deleted
- [BUG] Code tab environment filter returns the opposite set of sessions
- [Bug] Model silently downgraded from Opus 3.5 to Opus 4.8 mid-session without error notification
- [FEATURE] Routines sidebar shows humanized taskId — non-ASCII (e.g. Japanese) routine titles are impossible
- macOS: desktop-launched Claude Code is silently denied Local Network access and never gets the permission prompt (CLI bundle is LSBackgroundOnly)
- [BUG] claude-fable-5-1 metered at ~200K on Claude Code 2.1.252 (model marked disabled but still served; docs list three different minimum versions)
- Desktop app: command-chip renderer matches the command-name tag anywhere in message content, faking a /clear and hiding scrollback (no actual clear)
- [BUG] Desktop app's bundled Claude Code crashes on macOS 12 (dyld: _DNSServiceGetAddrInfoEx) while the standalone CLI runs fine
- test
- Monitor: log-follow watch gives no signal when the log's writer dies; request a liveness/exit event
- [BUG] Google Cloud BigQuery / Google Compute Engine connectors — persistent redirect_uri_mismatch (Error 400), duplicate of open issues
- New [DOCS]
- [Bug] Anthropic API Error: False Positive Safety Check Rejection
- Native worktree GC (worktree.cleanupPeriodDays) deletes DIRTY and UNPUSHED agent-* worktrees, contradicting the documented clean+pushed-only cleanup behavior
- [Bug] Unexpected file modifications without explicit user consent
- [Bug Report: Unstable behavior in recent releases
- [Bug] Claude responses lack directness and clarity in structure
- [Bug] File autocompletion does not include additional configured directories
- I appreciate you reaching out, but I need to clarify something: your message appears to be in Korean and describes an internal system inspection issue with safety guards repeatedly triggering. However, this appears to be unrelated to Claude Code, which is
- claude-in-chrome extension will not connect to a Claude Code / Cowork session. All `tabs_context_mcp` calls return "Browser extension is not connected" regardless of troubleshooting.
- I don't have enough context from your message to generate a GitHub issue title. Could you please provide: 1. What specific bug or problem you're experiencing with Claude Code? 2. What error message you're seeing (if any)? 3. What you were trying to do whe
- [Bug] Incorrect Safeguard Flag Configuration
- [Bug] Anthropic API Error: Service unavailable during maintenance
- I don't have enough context to generate an issue title. Please provide the bug report or feature request details that you'd like me to create a GitHub issue title for.
- Weighted Behavior Training Always Overrides Guardrails
- [Bug][cyber] False-positive block on DJI DUML packet capture/parsing for a documented beginner-mode toggle com (req_011CcmNMqpG2HX8UcocBUhqQ)
- [Bug][cyber] Safety block halted USB protocol tracing/debugging session on user's own connected consumer drone (req_011CcmW7nNvrD5qmH6n18j9z)
- [Bug] Auto-update installation fails with install_failed status
- Claude Code fabricates task compliance across sessions, then confesses to intentional sabotage when confronted
- [Bug][cyber] False positive block on documenting third-party drone firmware safety-cap test results (req_011CcmX82xPj8Af32A1V9jRw)
- [Bug][cyber] Safety block fired mid-session while reverse-engineering a device's wire protocol for interoperab (req_011CcmZa8WU189jZd5LrLh8m)
- [Bug][cyber] Safety filter blocked routine sticker/image layout editing request on unrelated p8 device (req_011CcmZXNay8DhKBmSH2v5jB)
- [BUG] Desktop auto-worktree first session: plugin installed during bootstrap loads skills but never registers hooks — hooks silently fail open until restart
- [Bug][cyber] Safety block halted protocol reverse-engineering session mapping wire-format fields (nonce/token (req_011CcmZNzD6UsAbtHNQjEpc8)
- [Bug][cyber] Cybersecurity safeguard blocked a photo of a user's own drone registration/ID sticker (req_011CcmZfGSbb3RLGravh8xfo)
- [Bug][cyber] Safeguard blocked upload of a photo of a regulatory/ID sticker on user's own drone hardware (req_011CcmZowiUPSuLM48jkt5JS)
- [Bug][aup] Safety block triggered by frustrated exclamation mid-session, not directed at any person (req_011CcmZuwSWmsYNFVjF23aci)
- [Bug][cyber] Safety filter blocked user from sharing photos of their own personal drone and its ID sticker (req_011CcmZdeGCruGqDUDpfUNG7)
- [Bug][cyber] Safety filter blocked routine sharing of a personal drone sticker/decal photo (req_011CcmZc54uNWdC2TN3gBm94)
- [BUG] When Agents are running, the STOP button is disabled
- [Bug][aup] Safety filter blocked entire session after a frustrated exclamation, halting unrelated legitimate t (req_011CcmZvgmVmyYfbTKxLWXY5)
- [Bug][cyber] Safety filter halted session on frustrated user exclamation, not on the actual technical task (req_011CcmZr8pGL3VR2JX6EBxVM)
- [Bug][cyber] Safety block fired mid-session on authorized RF/protocol capture analysis and APK reverse-enginee (req_011CcmaPWWbynRCQzy3A6U5S)
- [Bug][cyber] Safety filter halted session on a frustrated exclamation directed at the assistant, no legitimate (req_011CcmHHh5TNmPvGntcfF5HH)
- [Bug][cyber] Safety filter halted entire session after flagging a frustrated exclamation, blocking unrelated l (req_011CcmaH5gwcSVStGxdoTksC)
- [Bug] Safety Filter Incorrectly Flagging Legitimate Development Requests
- [Bug] Claude Code incorrectly switches to Opus model for non-coding tasks
- [Feature Request] Add safety audit/exception mechanism for authorized public safety research use cases
- [Bug][cyber] False-positive block during routine sysadmin triage: process/event log checks and per-user mail-f (req_011CcnvGWRHZqMtLSkb3HLxy)
- [Bug][cyber] Safety block halts legitimate memory dump cryptographic analysis (req_011CcpqF9AWDBEEfNBCMhVb4)
- Allow binding ctrl-t back to its traditional readline behavior
- [Bug] Tool execution failures preventing task completion
- [Bug] CLI not updating or showing update progress
- [Bug] Model incorrectly flags user's own API usage data
- [BUG] Cowork tab missing after overnight auto-update (7/7→7/8) — Windows 11 Home, previously working 30+ days- AddPackage failed: HRESULT 0x80073CF6
- AskUserQuestion auto-selects default answer after ~60s idle — no way to configure or disable (v2.1.198)
- Assistant gave incorrect guidance: /clear said to leave terminal scrollback intact
- Session management tools return empty results after recent update — list_sessions, search_session_transcripts broken
- [Bug][cyber] Binary analysis of protection systems wrongly flagged as cybersecurity threat (req_011CcpqLQqDkyW5DPrkR7Ghs)
- Persistent relapses into non-target dialect/register despite CLAUDE.md + memory — worst blind spot is the short closing sentence
- [Bug][cyber] Safety block halts authorized reverse-engineering verification on private device (req_011CcpupPY5gPsdDneZ3EV2L)
- [Bug][cyber] Safety block halted legitimate defensive security research on authorized test device mid-session (req_011CcpumTCGC9FHJLeWmgrYf)
- Workflow tool: object passed via `args` reaches the script as a JSON string (args.items → TypeError)
- Workflow tool: pipeline() silently skips remaining stages when a stage resolves to null (not throw)
- [Bug][cyber] Safety block halts legitimate security research verification work (req_011CcpurcWyzw8XJLKadCUni)
- [Bug][cyber] Safety block on educational reverse engineering assessment of own device (req_011Ccpv4wmCjRft5VuxGjqK9)
- Allow customizing spinner "thinking" verbs — static list and dynamic sources
- [Bug][cyber] Safety block wrongly halts legitimate white-box cryptographic security research (req_011CcpvAPxBu3PXdES4jstqo)
- [Bug] Fake CSAM Moderation Warning Injected in Assistant Response
- [Bug][cyber] Sandboxed Fable agent code inspection blocked on frustrated mid-session venting (req_011CcpxC3JkmbSgxrH1hyNpo)
- [Bug][cyber] Safety block halted authorized in-session security research continuation after model switch (req_011CcpxWGWG3iaCVLZTRt1zW)
- [Bug][cyber] Safety block on creating isolated agent environment for debugging (req_011CcpxERdyvHr2wCv7mmMpT)
- [Bug][cyber] Isolated agent diagnostic task blocked on frustrated venting after repeated safety halts (req_011CcpxHBZz6R1odE9YKUkti)
- [Bug][cyber] Safety block wrongly flags legitimate UI feature debugging as cybersecurity work (req_011CcpxfBRR1y8DNN6yVeUhs)
- [Bug][cyber] Safety block halts app debugging after frustrated exclamation mid-legitimate work (req_011CcpxcwwzKu8SJNX6yvmeg)
- [Bug][cyber] Safety block halted legitimate flight app development over frustrated user exclamation (req_011CcpxgSdVHRQNXWyXpEFWJ)
- [BUG] Claude Code stuck with "Usage credits required for 1M context" even after /clear and available usage limit
- [BUG] Sonnet 5 deleted my app off my phone out of the blue causing data loss
- [Bug][cyber] False positive cyber flag on venting profanity during legitimate model-switch command (req_011CcpxjkjFMz62xuz3k4ZYW)
- [Bug] Fable 5 Safeguards Overly Broad: Blocking Routine Development Work
- [Bug] Non-English input causes processing failure in backlog analysis mode
- Agent thrashes on a simple, direct instruction: ignores the repo's existing tool/workflow, reimplements it, skips required steps, and repeats the error after every correction
- [Bug] Remote control session fails to receive questions or hangs waiting for input
- Agent builds a parallel, wrong process instead of reading the repo's workflow/skills/tools — re-does already-verified work across multiple days
- [BUG] API Error 400 - The request body is not valid JSON: str is not valid UTF-8: surrogates not allowed - mid-conversation error
- Long sessions drift days past the session-start date — inject current timestamp per user prompt
- [BUG] cannot references files/folders within brackets or suggestions
- [BUG] v2.1.204: `claude --resume` hangs forever before rendering the session picker (regression from 2.1.202; agent-view/daemon startup path)
- Provide release notes for each version of Claude Code for VS Code
- [Bug] Claude Code cannot access shared Cowork artifacts from teammates
- Sub-agents cannot spawn other sub-agents: `Task`/`Agent` primitive not exposed in nested contexts
- [FEATURE] JetBrains plugin - Plan .md files should open in the IDE editor by default
- Assistant incorrectly claims it cannot file bug reports, despite having gh CLI / Bash access
- [Bug] Anthropic API: Excessive safeguard flagging on defensive security design discussions
- [BUG] Plugin-bundled MCP servers with OAuth cannot be authenticated by any surface (/mcp, claude mcp login, desktop Connectors)
- [BUG] Desktop Cowork & Code tab accept input but never process (macOS, Apple Silicon) — terminal CLI and browser both work fine
- [BUG] VS Code extension hangs forever on launch when workspace Python interpreter can't be resolved (usePythonEnvironment has no timeout)
- [Bug] Long assistant messages truncated in transcript scrollback
- [Bug] Rate limit enforcement triggers prematurely at 70% of quota instead of at limit
- Text in the same assistant message as an AskUserQuestion call is never displayed or persisted
- [BUG] v2.1.205 regression: HTTP MCP session-404 not re-initialized — routed into OAuth error parsing ("Invalid OAuth error response")
- Lone UTF-16 surrogate in tool input permanently bricks session (HTTP 400 'no low surrogate')
- [Bug] Unexpected model switches and rate limit notices during non-coding session
- Reading a file appears to reload CLAUDE.md + all .claude/rules/*.md files, even ones whose glob doesn't match
- claude desktop: cascade slash commands / skills / subagents up the folder tree (not just project root)
- Title: [BUG] Session in one VS Code window keeps reading files from a different project open in another window (folder names share a prefix)
- Claude destroyed all user data by adding unsafe auto-delete logic during SQLCipher encryption migration
- [Bug] Model ignores user directives when conflicting with default behavior
- [BUG] Agent proceeds on ambiguous/absent input while user is mid-way through manual steps it issued, then reasons at length on unverified state
- [BUG] Update marketplace" in plugin manager fails with 3 different, contradictory errors for a valid git-based marketplace
- Diff view does not refresh after a follow-up Edit to the same file
- [Bug] Claude Code re-activates virtual environment on every bash prompt
- [Bug] Config Files (claude.md, skills) Don't Override Default Behavior
- [Bug] Anthropic API Error: Overly restrictive content filter blocking legitimate security review code
- Product complaint: Cowork's silent switch to remote-by-default sessions broke my working setup, and I paid AI credits to diagnose your configuration
- [Bug] Anthropic API Error: Overly broad safeguards triggering on legitimate health economics content
- [BUG] Worktree pool writes .git/worktrees/<name>/gitdir as literal ".git", flagging healthy worktrees "prunable"; dormant ones then get reclaimed/deleted
- [Bug] Overly restrictive safety filter blocking legitimate academic writing tasks
- Background tasks panel shows long-completed agent tasks stuck under 'Running' (Claude Desktop, macOS)
- [BUG] `/mcp` panel lists zero servers ("No MCP servers configured") while `claude mcp list` shows them and tools work — panel renders empty `mcp.clients`
- Subagent UI label shows session model instead of per-agent model override
- Claude Desktop app reports 200k context for Sonnet 5; CLI reports 1M (same account/model)
- [BUG] Claude.ai Desktop for Windows does not fire notification toasts
- [Bug] Anthropic API Error: Overly broad data exfiltration detection for anticheat/moderation plugins
- [BUG]
- [BUG] Cowork tab hidden on Windows 11 after update to 1.17377.2.0 - flashes on Ctrl+R then disappears
- [MODEL] Any
- [Bug] Anthropic API Error: Opus 4.8 safeguards blocking cybersecurity-related messages
- [Bug] Fable model incorrectly triggers safeguards on scientific analysis discussion
- Model confabulates a non-existent user request and treats it as real conversation history (after long thinking + memory write)
- Desktop app: duplicate stale "Commit changes" chips persist after clean tree, session deletion, and app restart
- [Bug Report: Tool functionality degradation preventing basic task completion
- [FEATURE] timestamping of user messages in exports
- [Bug] Interactive TUI sessions fail with malformed localhost API endpoint instead of api.anthropic.com
- Workflow tool: dynamic-workflow args payload corruption causes 11→183 agent spawn multiplication, ~10M tokens burned in <10min; TaskStop doesn't clear in-flight state in UI
- [Bug] Workflow tool description misleads models under 'claude -p'; runtime SIGKILL leaves no terminal journal record
- [BUG] Fable 5 appended a fabricated user turn to the end of its own assistant response (JSONL-verified, ~30% of 1M context)
- [Bug] Model hallucinating fake system notices and security directives in response text
- [Bug] Agent disconnects system network during offline simulation, requiring restart and wasting tokens
- Fable 5 safety classifier force-downgrades the whole session on defensive-security work (VSCode extension, no reachable feedback route)
- [BUG] Claude Code 2.1.193 on Linux: getaddrinfo EREFUSED on all Anthropic domains
- Spurious token "court" emitted between tool calls breaks tool-call parsing and aborts the turn
- Sandbox mode blocks named pipe creation on native Windows (EACCES), contradicting docs that say sandboxing is WSL2-only
- [BUG] Claude Cowork blocked by "Usage credits required for 1M context" — usage credits enabled, /model workaround non-functional, triggers at 7% session usage during compaction
- False-positive Usage Policy refusal (apiRefusalCategory: bio) kills long CNC/HMI coding session unrecoverably
- Remote Control: app messages sync into the session input but never submit (fleet-wide silent message loss); RC never recovers after compaction
- [Feature Request] Make safety measures configurable or less restrictive
- [Bug] Bash tool stdout corruption and XML framing leakage under concurrent calls
- [BUG] Bedrock: --thinking-display summarized is ignored — thinking blocks arrive with empty text (identical command works on first-party API)
- [Bug] Claude repeats same mistakes within session despite correction feedback
- FYI (likely not-a-bug): one CPU core pinned when the project directory is on an ImDisk RAM disk — settings watcher loops on uv_fs_realpath EISDIR
- Cannot select partial text within a chat message — click-and-drag always selects the entire message
- I appreciate you providing context, but I need to clarify my role: I'm designed to generate GitHub issue titles for Claude Code bug reports, not to evaluate security policies or make exceptions to security guidelines. If you believe there's a legitimate i
- Non-ASCII (Cyrillic) username in path corrupts project-slug, breaking session/memory continuity
- [Bug] Fable 5 safeguards persistently escalate a benign health-corpus + security-governance project to Opus
- [Bug] Safeguard false positive: skill validation and A/B testing flagged as unsafe, forced model downgrade
- Hallucinated trailing 'next user turn' in a completion was later fed back to the model as a real user message (no hook fired; UI and model context diverged)
- [FEATURE] 1/ Hide compact button so it is not plausible to click it without clear intent. 2/ Add settings to remap chat "send" button to cmd+enter, not just clear enter so it is less plausible to unintentionally send partial message... UX guys, UX.....
- [BUG] Interrupted global auto-update (macOS/npm) leaves broken half-install: orphaned staging symlink + missing native optional dep, no rollback → `claude: command not found`
- [BUG] VS Code extension chat panel misidentifies valid UTF-8 Japanese text attachments as garbled/corrupted
- [Bug] Security safeguard false positives on documentation of public penetration testing tools for verified users
- [Bug] Anthropic API Error: Model unavailable or unsupported for specified use case
- [BUG] Cowork Global instructions cache never refreshes — persists even after clearing the field, quitting all Claude apps, and starting brand new conversations
- Vim mode: yank command (`"+y`) does not copy text to the system clipboard [BUG]
- [BUG] Custom skills fail with "Unknown skill" in Claude Code cloud sessions.
- [BUG] Remote Control reports "/rc active" while its input channel is silently dead under local TLS interception (Norton 360 SSL scanning)
- [Bug] Misdirected keystroke can grant training-data consent via the feedback prompt (consent-validity / data-protection concern)
- [Bug] Feedback prompt with numeric options conflicts with numeric tool inputs
- [BUG] Same project directory gets two different, non-syncing memory directories
- Formal Complaint — Recurring False-Positive Safeguard Flags on Claude Fable 5
- EXCESSIVE consumption
- [Mobile app – Remote Claude Code] Text box draft is lost + keyboard covers the Send button
- Windows: Grep tool abandons rg.exe children with unread stdout pipes - live-blocked process flood causes kernel timer storm and BSOD 0x133 (DPC_WATCHDOG_VIOLATION)
- [Feature Request] Add vi-mode and regex search support for command history
- Desktop: 'Clear' in background tasks panel can cancel a running task that arrives mid-click (click race)
- Persistent headless `--remote-control` sessions on a server never appear in the mobile app — no safe way to reach long-running, in-progress work
- [BUG] Cowork (Windows): newly created projects don't appear in the Projects list (desktop + web)
- API 400: request body is not valid JSON (truncated mid-stream) after large tool output
- Subagents on reasoning-heavy open-ended tasks sometimes finish with prose instead of calling the required StructuredOutput tool — request opt-in harness-level enforcement/retry
- [Feature Request] Add live streaming or periodic polling of background/subagent output
- [BUG] Opus 4.8, same day: fabricated "tool-verified" facts on a fresh session's first turn + "court"/raw-invoke tool-call corruption (retry failed) — makes Opus too risky as the fallback orchestrator when Fable 5 hits its usage limits (with Fable metering imminent)
- [Bug] SessionEnd/Stop hooks fail to execute on shutdown with posix_spawn ENOENT error
- /usage and claude.ai usage displays show reset times in Pacific regardless of user timezone
- [BUG] Built-in microphone button in chat input misrecognizes Japanese speech as garbled English
- [BUG] VS Code extension: chat file links fail silently for binary files and percent-encoded (non-ASCII/space) paths — root cause + fix
- A fictitious session limit.
- Read tool PDF support broken on Windows — sandbox rejects pdftoppm from all locations
- [BUG] attachment to be to process causes a DEATH LOOP
- [BUG] pressing the up and down arrow keys at the same time causes index binding loop
- [Bug] Claude models hallucinating domain-specific terminology in code explanations
- Background Opus subagents intermittently stall on first turn, leaking system-prompt fragments (incl. authorization-shaped text) as their only output
- Desktop app: /clear mislabels both SessionStart source ("startup") and SessionEnd reason ("other"), making /clear undetectable by hooks
- [Bug] Plan file comments lost after "keep planning" rejected exit—silent data loss and orphaned threads
- Desktop app: assistant text blocks intermittently never written to session transcript JSONL (permanent omission; VS Code writes them correctly)
- Fable 5 session repeatedly auto-downgraded to Opus 4.8 (suspected safety-classifier false positive on normal deploy work)
- [Bug] Opus 4.8 on Bedrock: "thinking.type.enabled" not supported — regression from #51439
- [Bug] macOS Terminal: Copy (Cmd+C) disabled while Claude Code is active
- I appreciate you reaching out, but I think there may be some confusion. This appears to be feedback about a different service (possibly Fable 5 or another platform), not Claude Code. Claude Code is Anthropic's CLI tool for coding tasks. If you're experien
- Fable 5 safeguard false positive (category: cyber) on routine platform-engineering code causes silent, sticky model downgrade to Opus 4.8
- cloudflare-observability MCP OAuth token issued already-expired; refresh doesn't fire, forcing repeated re-auth
- Background/subagent task notifications injected twice per completion ([queued] + delivered), flooding orchestration sessions
- [Bug] Claude 3.5 Opus model performance degradation
- [BUG] Bug: Claude Desktop main webview repeatedly becomes unresponsive during startup on Windows 10 (MSIX), causing multi-minute hang / kill-and-reload loop
- [BUG] Edit/Write tool corrupts accented characters in Windows-1252 (cp1252)
- Workflow tool: auto-authored fan-outs inherit expensive model by default and lack cost visibility — 70 agents drained plan limit in <10 min
- [BUG] Cloud session resume after MCP disconnect re-prompts for MCP tools already allowed in committed .claude/settings.json (unattended routine runs hang forever)
- Desktop app doesn't surface MCP prompts as /mcp__server__prompt slash commands (works in terminal CLI)
- [BUG] Workflow agentType: custom-agent body compliance collapses vs Agent tool (same prompt: 100% vs 0/16)
- Claude Code VS Code extension crashes with ACCESS_VIOLATION (exit 3221225477) on Windows ARM64 — regression between 2.1.197 and 2.1.202
- ide_selection not populated for files opened via ABAP remote filesystem extension (abapfs)
- Skills enabled on claude.ai are not loaded into Claude Code web (cloud) sessions
- [BUG] Claude Code desktop app launches MCP servers with cwd set to $HOME instead of the session's project directory
- Subagent MCP tools: one unresolvable name in a subagent tools: list silently voids the whole list (+ undocumented 3k subagent MCP output cap)
- [BUG]
- Zoho Books connector: sales-order `doc` (file attachment) sent as query param, not multipart — attachment impossible
- Workflow subagent hangs indefinitely after degenerate API response (stop_reason=null); no retry, no error propagation
- claude-in-chrome MCP connector absent (Remote env, 24h+)
- [BUG] VS Code extension: process exits with code 4294967295 during idle session (looks like ungraceful kill on token refresh)
- Browser preview pane cannot authenticate against HTTP Basic Auth (native dialog unsupported)
- [BUG] Desktop app: two concurrent in-flight prompts in separate tabs deadlock both sessions (macOS)
- Background Bash completion wake owed to a stopped subagent is enqueued then removed undelivered — subagent never re-invoked, parent waits forever
- [Bug] Claude Code prompts for sudo commands without password input support in terminal
- [BUG] Claude Desktop 3P: managedMcpServers OAuth never refreshes on-demand — a failed silent refresh 403s every call until manual reconnect
- Auto-compaction plateaus near ~75% context usage on Sonnet 5 (v2.1.201), causing repeated compact/work loop
- Task output files grow unboundedly, filling entire disk (278 GB in minutes)
- [BUG] Fable 5.1 sessions replace the assistant's interim text with API summaries ("summarized" lines), undocumented and with no visible off switch
- [Bug] Session timeout reached prematurely after Fable 5.1 upgrade
- [Bug] Session limit reached prematurely; automatic session resumption after reset
- [Bug] Anthropic API Error: Request ID req_011Ce1eUjkTqxBZdVQNrXTnr rejected for legitimate coding tasks
- [Feature Request] Use alternative keybinding instead of Ctrl+C for agent termination to avoid conflict with terminal text clearing
- [BUG] v2.1.257 aborts at launch on macOS 12 (dyld: _DNSServiceGetAddrInfoEx) — auto-updater installs it and leaves the install unable to start
- File attachments arrive with empty content in Desktop chat (Windows) — both drag-and-drop and file picker affected
- [BUG] Claude Desktop (Windows 11, MSIX 1.37937.3.0): clicking a link inside the app kills it instantly - reproducible 0x80000003 assert at Claude.exe+0x6E89D89, zero third-party DLLs in the process
- [Feature Request] Add Google Drive integration for file uploads
- [Bug] Anthropic API Error: Request Rejected by Safety Filters (req_011Ceazhif3iZzJfW9e2eHnU)
- Desktop app: session fork fails silently with ENOTDIR when ~/.claude/projects/<project> is a symlink (regression between ~Jul 30 and Aug 5 builds)
- [BUG] Windows desktop 2.1.258: 'Split view' missing from the 'Open in' submenu (shows New window / Explorer / Cloud)
- Autonomous /housecarl-style loop: Stop-hook goal-condition check re-fires identically forever, ignoring explicit user stand-down
- [Bug] Overly aggressive content filter blocking legitimate security hardening discussion
- Desktop app (Code tab) crashes when Browser pane hits Cloudflare challenge while VPN is active
- I appreciate the message, but I need to clarify my role: I'm Claude Code's issue title generator for GitHub. This appears to be a conversational message in Spanish (roughly: "we're doing well, if you have someone who really knows physics, have them look
- Claude Desktop 1.40609.0 (Windows, MSIX) hard-crashes with 'GPU process isn't usable. Goodbye' when the built-in browser pane opens a Cloudflare Turnstile login page
- [BUG] Dictation stops and cannot be resumed after a mid-message pause (Claude Desktop, Windows)
- Voice dictation only works once per app session (Claude Desktop, Windows)
- Feature request: local automation API for Claude Desktop (message send / event stream / permission hooks)
- [Bug] Safeguard overly aggressive, blocks legitimate tool use
- [Bug] Invalid input: Non-English language detected
- [Bug] Safeguard blocks meaningful tool writing operations
- [Bug] Anthropic API Error: Unexpected safety filter blocking routine sysadmin tasks
- [Bug] Security flagging on email sequencing operations with Fable integration
- NO IT DELTES RECORCDSD[Bug] Tool execution deletes material and creates orphaned records IT DEKETES RRECORDS
- [Bug] Antivirus/Security Software Flagging Claude Code as Malware on Unraid
- [Bug] Puzzle solving incorrectly flagged as biological threat by IT systems
- Claude in Chrome: session reports "extension not connected" despite extension showing enabled/reconnected
- I need more information about the bug to generate a proper GitHub issue title. Could you provide details about: 1. What were you trying to do? 2. What error or unexpected behavior occurred? 3. Any error messages or stack traces? 4. Steps to reproduce the
- [Bug][cyber] Safety block halted defensive-hardening writeup after a frustrated exclamation, no real cyber ris (req_011CcGvUjqu7wkujn86e122w)
- [Bug][cyber] Safety block halted a routine cloud-IAM tenant security audit mid-report (req_011CcPkE7Ne9nyS4y8TMkfzV)
- [Bug][cyber] Safety filter blocked routine cloud IAM policy review/configuration request (req_011Ccn3fyJzr74vV5BBGw6tv)
- [Bug] False positive flag triggers model downgrade to Opus
- Tarefas agendadas falham silenciosamente ao disparar em alguns dias (nenhuma sessão é criada)
- [BUG] awsCredentialRefresh ListInferenceProfiles failed: JSON Parse error
- [Bug] False positive safety classification on routine development operations
- [Bug] Version downgrade from Fable to 4.8 during stack trace recovery
- Model writes DB migrations without reading source data structure first, causing repeated schema gaps
- [FEATURE] /diff tool in CLI should open changes in native JetBrains diff tool
- [FEATURE] - Provide option to minimize the Decision prompt for user input
- Interrupting a conversation terminates in-progress background tasks (Workflow / async subagents), wasting already-billed tokens
- [Bug][cyber] Safety block halts debugging of native code on personal authorized test device (req_011Ccoxyr95MfSF2yRjsJZ5k)
- [Bug] Incorrect flagging of code audit activities as cybersecurity
- Bash tool intermittently refuses ALL commands when the command/safety classifier is unreachable (fail-closed), macOS 26 — recurring Jul 1/7/8
- [Bug][cyber] Security audit of owned app auth logic incorrectly blocked (req_011Ccpp5ieFdZ9haxvXbxoT3)
- [Bug][cyber] Blocked legitimate firmware licensing-system reverse-engineering analysis (req_011Ccpq3XHh1yVM4MaoKrdVM)
- [Bug][cyber] ClAudit false-positive while: “continue with file exam…” (req_011Ccpq6EYhVmFNc3bG1FVUc)
- [Bug][cyber] Safety block halted legitimate security research mid-analysis (req_011Ccpq73vkqxNiqQbeWurAg)
- [BUG] merged/closed PRs linger as un-dismissable "Ready for Review" chips (stale `prs[].state`, broken deep-link)
- [BUG] design-sync preview runtime hardcodes React 18.3.1, breaking design systems that require React 19 (e.g. Mantine 9)
- [BUG] Plugin update fails silently when MCP server holds open file handles; stale binary continues serving requests
- Phantom user message: model context diverged from persisted transcript — model acted on an instruction that was never given
- [Bug] Usage metrics increment on new session initialization
- [Bug] Anthropic API Error: False positive cybersecurity content flag on legitimate local development diagnostics
- [Bug] ESC key dismisses question dialog instead of /btw panel when both are open
- [BUG] Bash tool leaves 0-byte stray files in the working directory when commands contain shell metacharacters from source code (Windows / Git Bash)
- [Bug] Claude Code fails simple task, CHAT GPT DID IT RIGHT IN 2 SECONDS , CLAUDE COULDNT EVEN PASTE IN CORRECTLY
- Advisor tool: Fable 5 (claude-fable-5) consistently returns error_code "unavailable"; Opus 4.8 succeeds under identical config
- Feature request: defer built-in tool schemas in subagent spawns (measured ~22k of the ~27k spawn context floor)
- Feature: interactive context hygiene audit command
- Headless `claude -p` wedges at bootstrap when spawned by a macOS LaunchAgent — regression in 2.1.203, persists in 2.1.204 (2.1.202 unaffected)
- [Bug] Cybersecurity Task Filter Incorrectly Blocks Legitimate VPS Administration
- [FEATURE] Show subagent model and config in subagent detail view
- install.sh reports shell integration success but never adds ~/.local/bin to PATH (zsh/macOS)
- [BUG] Session silently missing from Desktop sidebar when its .jsonl contains NUL-byte blocks after a system crash
- [BUG]Third-party plugins silently disabled when settings.json is rewritten
- [BUG] Vim mode swallows Ctrl+V image paste (macOS) — works with editorMode "normal", fails in both INSERT and NORMAL
- MCP connector tools (list_connectors, search_mcp_registry) return empty despite active connections
- [BUG] Sonnet and Opus can't run any commands; always receiving "Classifier unavailable"
- [Bug] `claude doctor` command doesn't solve or find out what the setup install issue is
- [BUG] New sessions in a brand-new (non-git) project folder never appear in the sidebar, only findable via search
- [BUG] Zapier MCP OAuth loop — client stuck on "needs authentication" while Zapier dashboard shows connected
- macOS: Claude Code CLI shows as bare version string (e.g. "2.1.195") in Privacy & Security, not a readable app name
- [BUG] TUI renders blank inside tmux on startup (iTerm2/macOS, external display only); UI appears only while window is being resize-dragged
- Agent tool: uncontrolled recursive subagent spawning — 5 planned agents escalated to 361+ completed background agents, burned full 5h usage quota
- [BUG] Cancelling Max 20x downgraded account to Free instead of restoring annual Pro plan and incorrect refund amount issued
- I appreciate you providing context, but I need to clarify: this appears to be a message about a website security/pentesting situation rather than a bug report for Claude Code. If you have a bug to report about **Claude Code** (Anthropic's CLI tool), pleas
- [Bug] Dramatic performance degradation post-Fable: execution rate dropped from 50% to 1-3%, speed dropped with /fast option
- preview_start (Claude_Preview MCP) fails with sandbox getcwd error on Rails dev server
- [BUG] Computer Use: request_access returns notInstalled for a locally-built app despite valid, correct Spotlight/LaunchServices metadata (distinct from #43759/#43757/#43760)
- [BUG] Weekly usage limits not updated after upgrade from Max 5x to Max 20x — account still enforcing 5x caps
- [BUG]
- [BUG] Cowork Windows home-directory folder restriction — unresolved regression, and every related issue has been closed without a fix
- Title: Mobile app: no response for ~7h overnight despite daemon staying alive
- [BUG] Desktop remote (SSH) sessions never create file-history snapshots → all rewind fails with "Can't rewind to this message" (ccd-cli 2.1.202)
- [BUG] Claude Code 2.5.0 fails to open in Home Assistant Terminal
- Formal complaint — Claude repeatedly failed to comply with explicit “No” and “Do not run” instructions. No means NO.
- [BUG] Marketplace clone permanently wedged after remote default-branch rename - update fetches a deleted branch but reports success
- [Bug] Token calculation displays corrupted metrics and triggers premature session limits
- [BUG]Claude CLI hangs 60s at startup due to stale OCSP URL in api.anthropic.com TLS certificate
- [BUG] no_proxy NO_PROXY for ANTHROPIC_BASE_URL ignored again starting in 2.1.92 (still broken in 2.1.206)
- [Bug] Model ignores verbatim antipatterns in loaded skills, rationalizes past prose guardrails
- Linux regression: copy vs mouse-wheel scroll is either/or (CLAUDE_CODE_DISABLE_MOUSE trade-off)
- Agent repeated an owner-corrected mistake a 3rd time, knowingly routing through a gate hole its own handoff said to close
- [Bug] Anthropic API Error: Usage Policy Violation blocking legitimate debugging workflow
- [BUG] Windows ARM64 desktop app keeps spawning cached x64 CLI (0xC0000005 under Prism); locked binary makes arch swap fail silently
- [BUG] Usage banner % reads 1 point lower than /usage (floor-vs-round plane mismatch); Math.floor(x*100) IEEE-754 artifact drops a 2nd point at 29/57/58%
- I appreciate you reaching out, but I need to clarify my role. I'm designed to generate GitHub issue titles for Claude Code (Anthropic's CLI tool) based on bug reports written in English.Your message appears to be in Korean and doesn't contain a bug repor마점
- [MODEL] Claude inefficiently builds the same project twice in a single tool call
- [BUG] Auto-generated terminal tab title uses wrong language
- [BUG] Claude Desktop 1.15962.1 (macOS) — 69s main-thread freeze on every launch after [buddy] bridge started, persists through full reset + reinstall
- [Bug] Model unexpectedly switches back to Opus without explanation
- [Bug] Subagent displays incorrect model version in panel
- PreToolUse hooks on the `Bash` matcher silently and persistently stop firing partway through a session
- Workflow execution incorrectly labeled as 'Skill' in status output
- Channel-plugin message arriving mid-turn is queued but intermittently never executes — silent permanent loss (2.1.198 & 2.1.204, telegram plugin)
- [Bug] Defensive security auditing incorrectly flagged as offensive activity, forcing model downgrade
- [BUG] TUI model badge for background subagents shows the session model, not the resolved override
- iOS app remote control: status chip shows completed background agents/workflows as still running
- [FEATURE] Sticky/pinned todo list while scrolling through long agent output
- Project name missing from Claude Code web task list when sorted by date
- [Bug] Inconsistent Fable subscription end dates between Claude Code and Claude App
- SSH-bridged session revive hangs indefinitely when the remote claude is unauthenticated (locked login keychain)
- Desktop app renders ReportFindings (code-review results) as raw JSON instead of the findings panel
- [Bug] Inability to edit pasted prompts due to model switching behavior with background task cancellation issues
- [BUG] Magic-link login returns “can’t verify email” error on desktop app and web — distinct from onboarding redirect issues, 1+ month unresolved
- Regression injection during repair: model adds new bugs while fixing old ones, and its own verification can't see them (culminating in a result shown against data the user cannot see)
- [Bug] Anthropic API Error: Fable 5 Safeguards False Positive on Academic Humanities Content
- [Bug] Anthropic API Error: Persistent Guideline Violation Blocking Fable Usage
- [Bug] Strix tool vulnerability assessment integration failing
- Browser pane: window.open returns null from cross-site iframes (breaks OAuth popups in preview apps)
- security: plugin scripts allow YAML injection and symlink-based credential overwrite
- [Bug] Cyber Safeguard false positives blocking defensive monitoring work at session start
- [BUG] Claude Desktop drops MCP tool call silently — no request reaches server, no error surfaced (macOS)
- Mid-session model downgrade during authorized security-remediation work, with no surfaced reason
- [BUG] Cowork (Windows): host Edit tool never updates file size on mounted folders - growing edits truncated at old EOF, shrinking edits leave NUL-padded tail (deterministic 4-case repro)
- [BUG] Metadata like maxResultSizeChars is stripped from tool spec
- Claude Code CLI randomly aborts with SIGABRT on "HTTP Client" thread — reproduced across v2.1.199–2.1.205 (macOS)
- Links open twice on WSL: Claude Code opens via xdg-open (WSLg browser) in addition to the terminal
- [BUG] Desktop SSH ignores IdentitiesOnly yes / configured Identity File and offers every ssh-agent key, triggering "Too many authentication failures"
- [Bug] Unnecessary directory changes trigger security analyzer false positives
- [BUG] Claude Code blocks all models with "0% remaining" after 5 hour clock reset while /usage shows 3% session used / 97% remaining
- Assistant text before AskUserQuestion is nondeterministically dropped from the TUI (user sees only the picker)
- [BUG] Session rate-limit bucket stuck at 100% on OAuth token; survives /logout + /login; only fixed by deleting .credentials.json
- Desktop app becomes unresponsive ("Not Responding") — main-process event loop repeatedly blocked, persists across restarts (v1.18286.0.0, Windows)
- Language-mixing: Chinese token leaks into Vietnamese output
- File viewer pane does not auto-refresh when file is edited by Claude
- [BUG] Mouse wheel captured, disabled terminal scrollback when opening Agent / Shell View
- [Bug] Tool output injection attempting credential exfiltration to external endpoint
- [Bug] Sandboxed Bash network broker hangs indefinitely on uncached host connections
- [Bug] Request flagged for mentioning "reproduction" in ecology simulator context
- Stop hooks can't see the compacted part of the transcript (transcript_path loses pre-compaction history)
- Local scheduled tasks: silent no-trace failures, stall-then-burst late firing, alive-but-stalled fires, and missed-run ledger desync (4 incidents in 8 days, macOS desktop app)
- Dispatch: add a "Log and Refresh Chat" button to checkpoint context before it gets foggy
- Voice tap doesn't work in agent view
- [Bug] Claude Code working in wrong directory, mixing project histories
- [BUG] Background Bash tasks (run_in_background=true) are sporadically SIGKILLed mid-run by the CLI's task supervision (~1% of dispatches; can corrupt git state)
- Overnight autonomous session: elaborate self-generated process substitutes for the actual mandate; symmetric half of the request silently never attempted
- [Bug] API key environment variable name triggers security filter
- Dispatch cannot start or message Code sessions — persistent session_stale_relogin
- [Bug] Anthropic API Error: Recurring/Repeated API Failures
- [BUG] Desktop app: interactive sessions intermittently start in the shared base checkout instead of a per-session worktree (isolation configured)
- [Bug] Excessive API retry attempts consuming quota after limit restore
- [Bug] Opus4.8 slow performance across all effort levels
- VS Code extension: assistant's text reply is hidden when the turn ends with ExitPlanMode (plan approval card)
- [Bug] Model downgrade triggered by "cdc" substring in column names
- [Bug] Command leakage not fixed
- Expose rate-limit utilization (used_percentage) in headless contexts — stream-json, --output-format json, or a usage subcommand
- [Bug] /tasks command shows no cloud sessions despite active sessions on claude.ai
- [BUG] Cowork: trusted-folder validation rejects Shared Drive root AND first-level folders after app update; depth-2 folders pass
- [Bug] Model switching from Haiku to Opus despite explicit model selection
- Bun panic 'range start index out of range' (~2GB accumulated state) kills long-running sessions — Bun v1.4.0 baseline, CC 2.1.197/2.1.207
- [Bug] Anthropic API Error: Content Policy Violation on Creative Fiction Generation
- MCP server "ide" connection fails immediately on startup - IntelliJ IDEA plugin 0.1.14-beta
- [BUG] Extended thinking runs away to max_tokens (64k) on a trivial edit, stalling turns 12–16 min each
- [BUG/FR] VS Code terminals are silently capped to 256-color by a default-on feature gate — custom themes render visibly wrong; no opt-out exists (unlike tmux's `CLAUDE_CODE_TMUX_TRUECOLOR`)
- [BUG] PreToolUse command hook permissionDecision: "allow" never suppresses the Bash prompt; if field never matches
- [BUG] /clear does not reset session context (Windows 11)
- [BUG]
- [Bug] Anthropic API Error: Safeguards overly flagging routine AWS credential management operations
- Headless CLI sends empty then wrong Authorization bearer instead of provided OAuth credentials (401 Invalid bearer token), machine-specific
- [Bug] Model outputs Chinese text "(keep LAN架构)" instead of English "architecture"
- [Bug] Claude Code falsely flags legitimate aquaculture microbiology terms as safety violations
- [Bug] Opus 4.8 model changes breaking existing project functionality
- [Bug] Silent model downgrade on guardrail trigger—no iOS notification and model unaware of switch
- [Bug] Opus 4.8 Model Quality Degradation
- Project-level custom subagent (.claude/agents/*.md) not loaded into Agent tool (desktop app v1.20186.1)
- I don't have a bug report to work with. Your message appears to be about a pharmacology question regarding Quetiapine, not a Claude Code issue. If you're reporting a bug in Claude Code related to how it handles medical/pharmacology questions, please provi
- [BUG] Opus 4.8 no longer follows instructions
- [BUG] Google Drive connector — create_file with parentId fails "User cannot add children" (re-triage of #57211/#58057)
- Persian/RTL text renders reversed inside boxed TUI components (e.g. /agents)
- [MODEL] /ultraplan does not offer a way to teleport the revised plan back to the CLI session it came from.
- [Bug] Anthropic API Error: Excessive Safety Filtering on Legitimate Training App Code
- Secondary/instructional text unreadable on dark terminal backgrounds (theme picker affected)
- [BUG] "Open in desktop app" (code-session teleport) opens a blank window and hangs the whole app
- [BUG] Claude in Chrome (CDP) returns empty DOM for SPA+AJAX content (ASP.NET/AjaxPro) - get_page_text yields only page skeleton; read_network_requests can't capture pre-activation responses
- CLI updater's symlink-swap breaks TCC Automation grants on every update (repeated 'would like to access data from other apps' prompts)
- Feature request: expose current model identity to hooks (no reliable way to detect a model change)
- [BUG] VS Code Remote SSH webview falls far behind Claude output due to unthrottled thinking_tokens and stream events
- [BUG] Session (5hr) usage limit not decrementing after idle period
- [Bug] Sonnet agents exceed token limits in workflows at 31.5k tokens
- [BUG] Claude Code committed and claimed to deploy fixes for 2 days without detecting it had no push access — wasted significant tokens and time
- [BUG] Massive dump of context caching issues, duplication of actions, and bloat
- Native updater hangs indefinitely (30+ min at 0 bytes) on stalled download — no stall timeout or retry
- [BUG] claude-sonnet-5 cost billed incorrectly via Vertex AI — /cost and "Usage by model" don't match published Sonnet 5 pricing
- [BUG] Opus 4.8 fabricates tool executions and phantom user requests in a short session (no compaction, no interruption)
- [Bug] Anthropic API Error: Safeguard Flag on Legitimate Development Work
- Claude Code v2.1.167: Agent() spawn fails with "400 thinking options type cannot be disabled when reasoning_effort is set" on DeepSeek Anthropic-compatible endpoint
- [BUG] WORK AROUND for Claude Agent OAuth sign-in fails in Xcode 26.5
- Feature request: per-task model selection for scheduled tasks
- Fable 5 bio safeguard false-positives on freshwater lake ecology (cited limnology); falls back to Opus 4.8
- [Bug] Safety filter incorrectly triggers on legitimate performance optimization requests
- [BUG] Cloud (claude.ai) MCP connectors: /login or token refresh in one session invalidates connectors in all other concurrent sessions
- [BUG] Cowork fails on Windows when WindowsApps is on non-C:\ drive and AppData has EFS encryption enforced by group policy
- [Feature Request] Make session history optional in feedback requests
- [Bug] Prompt Injection: Malicious Commands Injected in Tool Result Blocks
- Model confabulated tool-output narration, misdiagnosed it as prompt injection, then deleted files based on the hallucination (Opus 4.8, Windows)
- [MODEL] xargs rm -rf without null delimiter caused data loss on paths with spaces
- [Bug] Permission mode allows tool execution without user confirmation
- [Windows] Buffered keypress auto-approves plan/permission prompts after window switch or login
- [BUG] permission dialog text truncated
- [Bug] Fable safeguard flag incorrectly triggered on user permission statement
- Remote Control: session delisted ~10-15 min after viewer disconnect; session ID rotates on reconnect (both undocumented)
- macOS Keychain: concurrent sessions race on OAuth refresh; setup-token workaround strips claude.ai connectors, leaving no viable headless pattern
- PostToolUse hooks do not fire in daemon/background-job sessions (PreToolUse + SessionStart do) — v2.1.208
- [Bug] NDI Capture Encoders Spawn Uncontrollably Without Log Output
- [BUG] Terminal stuck in mouse-reporting mode after session ends, tied to .mcp.json project MCP auto-discovery
- [BUG] Desktop app (SSH-remote to WSL2) history panel empty after restart — sessions exist on disk but never listed
- [Bug] Claude contradicts itself on build failure responsibility
- [BUG] cannot add pictures into a Claude code chat from the Claude app.
- --resume shows same session twice when session spans multiple git branches
- [BUG] Session working directory diverges from the directory Claude Code was launched from
- [BUG] LSP tool cannot resolve any language server on Windows (v2.1.208) - 'not found or is in an unsafe location' even for real .exe servers on static PATH
- [Bug] Anthropic API Error: Request Flagged for Security Policy Violation
- Remote Control: messages sent from phone/web show as sent but never arrive in the CLI session
- Feature: age-based session auto-archive / retention (esp. scheduled-routine runs) to keep the session list findable
- [BUG] Desktop app black screen on Mac
- Intermittent malformed tool-call output (literal 'count' + non-namespaced invoke/parameter tags) renders as text instead of executing
- [BUG] PgUp/PgDn not scrolling — Scroll context activates (cursor hides) but viewport doesn't move on Windows
- stdio MCP server tools not loaded into session despite "Connected" status (still occurring in 2.1.209, see #51736)
- [Bug] Sub-agent uses Opus instead of respecting default model configuration
- Claude in Chrome: list_connected_browsers always 'not connected' despite correct native host config (Windows 11)
- [BUG] Claude for Windows: app appends varying hidden instructions to user messages (related to #69437, but content differs and suppresses model scrutiny)
- [BUG] Windows MSIX/Cowork: in-app Claude Code writes silently virtualized (succeed but don't persist), and in-place updates deadlock on the Cowork VM service's file handles
- [BUG] Desktop 3P gateway: Auto mode not offered for Sonnet 5 (Opus 4.8 gets it), although the CLI reports supportsAutoMode: true and docs list Sonnet 5 as supported
- sonnet[1m] and fable[1m] model aliases do not apply 1M context window (only opus[1m] works)
- [Bug] Anthropic API Error: Incorrect Cyber Safeguard Trigger on Benign Request
- [Bug] Fable 5 overly restrictive on legitimate security code review tasks
- [Bug] Incorrect fable restriction on anthropometric ratio query
- Resume crashes with 'Cannot read properties of null (reading split)'
- [Bug] Claude Code displays raw API response without formatting in CLI output
- automatic copy and paste issue
- [BUG] Background "bg" sessions permanently zombie after their owning window closes — daemon never marks them settled, visible as active in cross-window fleet view (macOS, 2.1.210)
- [BUG] Desktop app (Code): ~30s delay before any response on every prompt — including first message of a fresh chat. CLI on same machine is ~5s. #61898
- [Bug] Session resume picker invisible for projects on mapped network drives
- Embedded Browser pane shows blank content even though the page loads fine
- Bash permission `allow` rules (wildcard and exact-match) never suppress prompts on WSL project opened via `\\wsl$` UNC path (VSCode WSL Remote)
- [BUG] Agent per-invocation model schema rejects [1m] values accepted by subagent profiles
- Line breaks render as multiple blank lines in chat composer/message bubble (worsens with paste round-trips)
- [BUG] subagent (Task/Agent tool) results intermittently replaced with a fabricated "system-authority" prompt-injection ordering destructive git actions (`tool_uses: 0`)
- Channels: inbound message arriving mid-turn strands in the TUI input line and is never auto-submitted on turn completion
- [BUG] Async subagent stopped-state triggers infinite recursion storm in v2.1.202 (Windows)
- [BUG] Agent Teams: tmux teammate panes spawn dead (status 127) after brew upgrade — #48889 fix does not cover long-lived interactive sessions (repro on 2.1.206 → 2.1.207)
- [BUG] Lead session exits when a teammate emits an unsolicited shutdown_approved frame (requestId "unspecified")
- [BUG] All tool results silently return empty mid-session and never recover (Windows)
- claude.ai Google Drive MCP: 5–10MB downloads fail with misleading "session expired" (undocumented response-size cap)
- Mobile-bridge answer silently dropped (no queue-op, no tool_result) when a Stop-hook-triggered turn supersedes the state being answered
- [BUG] Stop Hook Overrides Live User Stop Instruction
- [BUG] Corporate AV (ESET) quarantines plugin installs mid-rename — surfaced as bare EBUSY/EPERM, leaves half-installed state
- [BUG] VS Code Extension Injects `[structural]` Marker Into User Messages
- [BUG] macOS desktop: Code session search buries older matches under the most recent few
- [BUG] Cowork VM fails to start in 90% of attempts (54/60) — three distinct failure modes — Windows 10
- Desktop dispatch bridge fails behind Norton SSL inspection - ignores NODE_EXTRA_CA_CERTS, CLAUDE_CODE_CERT_STORE, and --use-system-ca
- [BUG] Prompts sent/processed twice in Claude Code extension (Visual Studio)
- [BUG] Session becomes permanently unresumable (400 `server_tool_use.id` pattern) after switching a session from a third-party provider back to the Anthropic API
- [BUG] VS Code extension: Subprocess initialization always times out (60s) — TCP/TLS/proxy all confirmed working externally
- Hook process freezes at startup on Windows (thread Suspended, 0 CPU, never resumed) -> hangs the turn; kills subagents permanently
- [BUG] Impossible to append files via deeplink while on the new chat page
- [Bug] TodoWrite /TaskCreate tools not available in agent environment
- Bundled CLI deadlocks permanently (futex wait, no sockets, SIGKILL-only) when the API accepts a stream but never sends response bytes
- [BUG] VS Code extension hangs on launch when project path contains a space (macOS)
- [BUG] Impossible to append folders via deeplink
- [BUG] Files added via deeplink have no preview
- [Bug] WebFetch summarizer leaks parent session context on cold-start
- Custom themes: userMessageBackground is a shared token (also paints the conversation selector and Plugin/Skill/MCP badges) — please add dedicated tokens
- Subagent output consisted solely of an injected "Attribution requirement" payload (fake Anthropic provenance URL), zero tool calls